Specify persistent SQLite scan database in README; plan in TODO.md

scan will maintain a persistent database of file signatures
(default /var/lib/sfdupes/db.sqlite, overridable via
SFDUPES_DATABASE) that survives between runs; rescans hash only new
or changed files (mtime/size) and remove records for vanished files,
so scan can be cronned daily. report and trees will read the
database instead of a scan stream.

TODO.md

@@ -14,8 +14,41 @@
 
 # Next Step
 
-- convert Makefile targets to scripts-to-rule-them-all `script/`
-  entrypoints like the other managed repos
+- persistent scan database (branch `persistent-database`): `scan`
+  maintains a SQLite database that survives between runs so it can be
+  cronned daily; `report` and `trees` read the database instead of a
+  scan stream. Database path defaults to `/var/lib/sfdupes/db.sqlite`,
+  overridable via `SFDUPES_DATABASE`. Plan:
+    - [x] update `README.md` (the authoritative spec) for the database:
+          new Database section, revised `scan`/`report`/`trees` modes,
+          dependency list (`modernc.org/sqlite`, pure Go, cgo stays
+          disabled), exit codes, smoke test with incremental steps
+    - [ ] add `modernc.org/sqlite` (hash-pinned via `go.sum`)
+    - [ ] `db.go`: path resolution (`SFDUPES_DATABASE` env, default
+          `/var/lib/sfdupes/db.sqlite`), open/create with WAL +
+          busy-timeout pragmas and `PRAGMA user_version` schema check,
+          `files` table (path BLOB primary key, size, mtime, head,
+          tail), load-all-rows, and single-transaction apply of
+          upserts/deletes
+    - [ ] `scan.go`: resolve operands to absolute paths; diff stat
+          results against loaded rows (reuse hashes when size matches
+          and mtime is not newer); hash only new/changed files; delete
+          rows under the scanned operands not successfully processed;
+          leave rows outside the operands untouched; new summary line
+          (added/updated/removed/unchanged/skipped); "update" progress
+          pass for the database write
+    - [ ] `report.go`/`trees.go`: read records from the database (no
+          positional args, no stream parsing); drop the NUL-stream
+          parser and malformed-record handling
+    - [ ] `main.go`: updated usage strings; `report`/`trees` take no
+          args (usage error otherwise)
+    - [ ] tests: db open/env-override/schema tests; incremental scan
+          tests (add, mtime update, delete, unchanged-hash-reuse,
+          out-of-scope rows untouched, error paths); rework pipeline
+          test to go through the database; keep walk/stat/hash/grouping
+          unit tests
+    - [ ] `.gitignore`: local `*.sqlite` artifacts
+    - [ ] `make check`, `make docker`, and the README smoke test pass
 
 # Completed Steps
 
@@ -33,6 +66,8 @@
 
 # Future Steps
 
+- convert Makefile targets to scripts-to-rule-them-all `script/`
+  entrypoints like the other managed repos
 - possible later features (explicitly out of scope per README):
   full-content verification of candidates, removal-script helpers