feat(import): import many games with background matching, review and undo

src/main/resources/db/migration/V22__Imports.sql

@@ -0,0 +1,51 @@
+-- Imports of many games at once (item 23): a list or a CSV, matched with IGDB in the background, then saved as
+-- games added by hand. A game an import added keeps its import_id, so the import can be undone.
+-- The table is also the queue of the two workers (GameImportWorker):
+--   QUEUED -> MATCHING -> READY (the user reviews) -> SAVING -> SAVED (-> UNDONE), or FAILED after 3 tries.
+-- queued_at orders each queue. progress_at is set while a worker runs the import and after every line; NULL while
+-- it waits. attempts counts the tries a watchdog or a restart stopped.
+CREATE TABLE game_import (
+    id          BIGSERIAL   PRIMARY KEY,
+    user_id     BIGINT      NOT NULL,
+    source      VARCHAR(16) NOT NULL,
+    status      VARCHAR(16) NOT NULL,
+    total       INT         NOT NULL,
+    created_at  TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+    queued_at   TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+    progress_at TIMESTAMPTZ,
+    attempts    INT         NOT NULL DEFAULT 0,
+    -- While SAVING: NULL = adding the games, DETAILS = loading their time to beat and moods.
+    step        VARCHAR(16),
+    saved_at    TIMESTAMPTZ
+);
+CREATE INDEX game_import_user ON game_import (user_id);
+CREATE INDEX game_import_queue ON game_import (status, queued_at);
+-- One import in the matching queue per user; an import of 1000 lines takes minutes of the shared IGDB budget.
+CREATE UNIQUE INDEX game_import_one_matching ON game_import (user_id) WHERE status IN ('QUEUED', 'MATCHING');
+
+-- One row per line of the input. igdb_id: the game the user already picked in the search (source SEARCH).
+-- matched_at NULL: not matched yet. The chosen_* columns hold the user's review, save_result what the save did.
+CREATE TABLE game_import_row (
+    import_id        BIGINT       NOT NULL REFERENCES game_import (id) ON DELETE CASCADE,
+    line_no          INT          NOT NULL,
+    title            VARCHAR(200),
+    platform         VARCHAR(32),
+    playtime_minutes INT,
+    status           VARCHAR(16),
+    igdb_id          BIGINT,
+    matched_at       TIMESTAMPTZ,
+    search_failed    BOOLEAN      NOT NULL DEFAULT FALSE,
+    exact            BOOLEAN      NOT NULL DEFAULT FALSE,
+    -- The IGDB hits, best first, kept whole (IgdbApiService.Game), so the save needs no IGDB call.
+    candidates       JSONB        NOT NULL DEFAULT '[]',
+    chosen_igdb_id   BIGINT,
+    chosen_platform  VARCHAR(32),
+    chosen_status    VARCHAR(16),
+    chosen_minutes   INT,
+    -- ADDED or SKIPPED (in the library already, or IGDB does not know the game).
+    save_result      VARCHAR(8),
+    PRIMARY KEY (import_id, line_no)
+);
+
+ALTER TABLE user_game ADD COLUMN import_id BIGINT REFERENCES game_import (id) ON DELETE SET NULL;
+CREATE INDEX user_game_import ON user_game (import_id) WHERE import_id IS NOT NULL;