chore(sqlite-out): the retired engine's name reaches zero (20 files -> 0), and two committed databases go with it

[ywatanabe1989]

Why

We do not use the retired embedded engine. Runtime state lives in PostgreSQL on
55432. Its name nevertheless survived in 20 tracked files here — including two
live database files committed into the tree.

Standing ruling applied: nothing outside docs/adr/ may keep the name, and
no guard test asserting its absence — so none is added.

Deleted outright

• .scitex/clew/runtime/db.sqlite and
  examples/_legacy/scitex/clew/multi_parent/scitex/clew.db — two real database
  files committed into the repository.
• examples/_legacy/notebooks/19_scitex_db.ipynb — a tutorial for the retired
  engine end to end: raw client connections, SELECT ... FROM sqlite_master
  introspection, "SQLite for local work" as the closing advice. Nothing survives
  removing the engine from it, so the notebook goes, and the master index's two
  links to it go with it.

Rewritten to name PostgreSQL

• README's scitex.db section (summary line, the worked example, the closing
  paragraph).
• docs/05_ADDITIONAL_MODULES.md module table row.
• The clew provenance prose in docs/sphinx/core_concepts.rst and
  docs/sphinx/modules/io.rst.
• examples/_legacy/scitex/io/schematic_io_concept.py and its recorded _out/
  YAML — the schematic's store box, kept consistent between source and output.

Tests

• tests/integration/test_integration.py no longer asserts the engine's wrapper
  class is exported from stx.db; it still asserts PostgreSQL and
  check_health.
• tests/integration/test_cross_package_imports.py drops
  scitex_io._load_modules._sqlite3. That module was deleted in scitex-io
  0.3.0, so the entry has silently skipped ever since — it was testing
  nothing.

Also

• The name leaves .gitignore (both copies of the Django block), the clew
  comment in pyproject.toml, the stdlib table in
  scripts/maintenance/_pypi_packages.py (behaviour-neutral: the name is not in
  KNOWN_PACKAGES either, so get_package_name still returns None), the
  CrossRef env example, and the legacy io-module guideline's format table.
• examples/_legacy/notebooks/27_scitex_scholar.ipynb no longer opens a raw
  client against the third-party impact-factor database. Scholar's own
  enrich_by_default=True already performs that enrichment — the removed blocks
  were labelled as fallbacks for when it did not.

A finding this PR cannot fix

Running the suite after deleting the two committed database files re-created
.scitex/clew/runtime/clew.db — 84 KB, first sixteen bytes SQLite format 3.
The writer is scitex-clew==0.17.0, an external pin; this repo cannot stop the
write. What it can stop is the file being committed again, so the second commit
adds an ignore rule for that path. Nothing was ignoring it before — the previous
file was tracked, so it never appeared as untracked noise and no rule was ever
added. The engine still runs at runtime via scitex-clew; that belongs in the
scitex-clew leaf, not here.

Verification

Positive control first — the same search, same root, for a string known to be
present, so the zero below cannot be a broken search:

$ git grep -il scitex HEAD | wc -l
2054

Then the actual query:

$ git grep -il sqlite HEAD
$ echo $?
1

Zero matches across the whole tree. docs/adr/ holds none in this repo either,
so no exemption is being leaned on.

561 passed, 101 skipped, 44 xfailed, 17 xpassed locally, 0 failed.

CI may still go red on the py3.13 leg for a reproducible segfault
(Fatal Python error: Segmentation fault, no JUnit report) that predates this
branch and is tracked separately — it is not caused by anything here.

Overlap

#362, #364, #365, #366 and #368 are open against the same subject. This branch
is cut fresh from origin/develop and takes the count to a measured zero; if it
lands, those should close.