feat(podcasts): concurrent download pool — fair share, tickets, admin tunables

Replaces the sequential worker with a download-manager-style pool
(§0 item 4). Model: the user's connection is the limit; semantics are
fair-share ACROSS feeds, not goroutine-per-book (a global FIFO would
make Book B wait behind Book A's 100 files).

- Atomic claim: GetNextQueuedEpisode + SetPodcastEpisodeDownloading
  was a read-then-write race under concurrency; ClaimQueuedPodcastEpisode
  converts queued→downloading in ONE guarded statement (two claimants,
  one winner — the loser gets ErrNoRows). SetPodcastEpisodeDownloading
  retired.
- Queue tickets: download_queue_pos BIGINT fed by a Postgres SEQUENCE
  stamped at enqueue — the enqueue loop's ordering IS the claim order
  (the foundation §0 item 5 builds on; previously all rows in a batch
  shared one updated_at, so within-feed order was arbitrary).
- Pool: up to podcast_download_concurrency (default 4, live range 1–16)
  claim-loop goroutines per drain round; the setting re-reads every
  round so admin changes take effect without restart.
- Fair share: claim prefers the feed least-recently-served
  (pickFeedToServe; never-served first), earliest ticket within a feed —
  books proceed in parallel while chapters land in order (start
  listening while the rest download).
- Bandwidth: shared token bucket across all workers
  (podcast_download_speed_mbps, default 0 = unlimited); take() reserves
  bandwidth and callers sleep outside the lock.
- Settings: new float setting type end to end (registry defaults +
  clamping, save validation, admin UI number-with-step input) + the two
  schema seeds; worker wires the registry via SetSettings (tests run on
  registered defaults).
- Tests: unit coverage for the bucket reservation math + fair-share
  pick; integration TestPodcastDownloadPool pins ticket order = enqueue
  order, claim exclusivity, and a full 3-episode pool drain.

Dev-DB note: the column/sequence/settings were applied live (idempotent
DDL, same statements the schema runs at next boot) since the running
container predates them.

internal/database/querier.go

@@ -1,6 +1,6 @@
 // Code generated by sqlc. DO NOT EDIT.
 // versions:
-//   sqlc v1.31.1
+//   sqlc v1.31.0
 
 package database
 
@@ -32,6 +32,11 @@ type Querier interface {
 	BulkUpdateFormatGroups(ctx context.Context) error
 	BulkUpdateProgressFromSync(ctx context.Context, arg BulkUpdateProgressFromSyncParams) ([]interface{}, error)
 	CheckForProgressConflicts(ctx context.Context, arg CheckForProgressConflictsParams) (int64, error)
+	// Atomic claim (§0 item 4): converts queued -> downloading in ONE
+	// statement. Two claimants racing on the same row: exactly one UPDATE
+	// matches download_state='queued'; the loser gets no rows (ErrNoRows).
+	// This is the SKIP-LOCKED guarantee without a second round-trip.
+	ClaimQueuedPodcastEpisode(ctx context.Context, id pgtype.UUID) (MediaItems, error)
 	// Cleanup expired OPDS tokens
 	CleanupExpiredOpdsTokens(ctx context.Context) error
 	CleanupExpiredRefreshTokens(ctx context.Context, dollar_1 float64) error
@@ -446,8 +451,6 @@ type Querier interface {
 	// GetMediaScanState reads one item's manifest (per-item rescan paths
 	// and the manifest integration tests).
 	GetMediaScanState(ctx context.Context, mediaItemID pgtype.UUID) (MediaScanState, error)
-	// Sequential download worker's claim source (single worker, day one).
-	GetNextQueuedEpisode(ctx context.Context) (MediaItems, error)
 	GetNextRetryTime(ctx context.Context) (interface{}, error)
 	GetNotStartedItems(ctx context.Context, arg GetNotStartedItemsParams) ([]MediaItems, error)
 	// Get OPDS token
@@ -656,6 +659,11 @@ type Querier interface {
 	// Item-side projections COALESCE to keep them non-nullable strings;
 	// the display file path falls back to the issue's own for row-less rows.
 	ListProcessingIssuesByLibrary(ctx context.Context, libraryID pgtype.UUID) ([]ListProcessingIssuesByLibraryRow, error)
+	// Fair-share candidate scan (§0 item 4): the download pool claims in
+	// TICKET order (download_queue_pos = SEQUENCE nextval at enqueue — the
+	// enqueue loop's ordering IS the download order, item 5). The worker
+	// groups these per feed and prefers the least-recently-served feed.
+	ListQueuedPodcastEpisodes(ctx context.Context) ([]ListQueuedPodcastEpisodesRow, error)
 	ListSyncConflictsByMediaItem(ctx context.Context, arg ListSyncConflictsByMediaItemParams) ([]SyncConflicts, error)
 	ListSyncConflictsByUser(ctx context.Context, userID pgtype.UUID) ([]ListSyncConflictsByUserRow, error)
 	// Download All + policy union worklist.
@@ -668,6 +676,9 @@ type Querier interface {
 	ListUsers(ctx context.Context) ([]ListUsersRow, error)
 	// First consecutive scan that cannot find the file on disk.
 	MarkMediaItemMissing(ctx context.Context, id pgtype.UUID) error
+	// SetPodcastEpisodeDownloading RETIRED (§0 item 4): the pair
+	// GetNextQueuedEpisode + SetPodcastEpisodeDownloading raced under
+	// concurrency; ClaimQueuedPodcastEpisode does both atomically.
 	// SUCCESS-only write (invariant 7): the file is on disk and probed.
 	// ffprobe facts (real duration via audio_metadata, precise size/mime)
 	// reconcile the feed's declared values.
@@ -753,10 +764,11 @@ type Querier interface {
 	SetDeviceSyncEnabled(ctx context.Context, arg SetDeviceSyncEnabledParams) (Devices, error)
 	// Library Visibility queries
 	SetLibraryVisibility(ctx context.Context, arg SetLibraryVisibilityParams) (LibraryVisibility, error)
-	// Worker claim: pending|failed -> queued. Attempts incremented at
-	// claim time so a crashed worker still counts the try.
+	// Enqueue: pending|failed -> queued, stamped with the NEXT QUEUE TICKET
+	// (§0 items 4/5) so claim order = enqueue order (the effective
+	// listening order the enqueue loop applied). Attempts increment at
+	// enqueue so a crashed worker still counts the try.
 	SetPodcastEpisodeDownloadQueued(ctx context.Context, id pgtype.UUID) (int64, error)
-	SetPodcastEpisodeDownloading(ctx context.Context, id pgtype.UUID) error
 	// Manual retry (same endpoint as download): reset to pending so the
 	// worker re-claims it; the reason text clears with the fresh attempt.
 	SetPodcastEpisodePending(ctx context.Context, id pgtype.UUID) error