SQLite tests: use the UTF-16 STAT4 allocation range

[brandonpayton]

Stacked on #783 so the same maintained Node and Chromium SQLite harness can exercise this test-only patch.

What #907 already landed

#907 landed the PHP batch and its platform fixes, including the mmap/munmap work. None of that runtime or ABI history is replayed here.

What remains here

SQLite's utf16 permutation stores the two STAT4 samples as roughly 2,000 bytes instead of the roughly 1,000 bytes used by the default encoding. This patch keeps analyze9-16.1's SCHEMA_USED assertion and selects the encoding-correct range when the active permutation is utf16.

What was deliberately removed

• no test is skipped or weakened;
• no Kandelo runtime, libc, VFS, or host behavior changes;
• no no_mutex_try, pager-fault, or other SQLite workaround patches from Fix mmap accounting under partial munmap churn #769; and
• no fake platform or hostname behavior.

ABI and package artifacts

There is no ABI change. The declared SQLite library output is unchanged, so this does not bump the SQLite package revision. Rebuilding the separate testfixture and generated browser test image does change those local validation artifacts, but neither is a committed package output in this PR.

Validation run

• the patch applies cleanly after SQLite: keep recursive tests within WebAssembly stack limits #780/SQLite: cap trigger recursion for WebAssembly stacks #781's five maintained SQLite patches;
• scripts/dev-shell.sh bash packages/registry/sqlite/build-testfixture.sh;
• Node all analyze9.test, which selected config=utf16: 147 cases, 0 errors;
• direct Chromium utf16 analyze9.test: 147 cases, 0 errors; and
• git diff --check.

Validation not run

• the full SQLite permutation inventory;
• full host Vitest and the non-slow browser suite;
• package staging, because the declared library bytes do not change; and
• performance benchmarks. This PR makes no performance claim.

A broader Chromium all analyze9.test diagnostic also ran the utf16 job successfully, but the overall command returned 1 because the separate pre-existing prepare.analyze9-9.4.1 job failed. That broader run is not claimed as passing validation for this PR.

[github-actions]

Phase B-1 matrix build status — pr-932-staging

ABI v39. 5 built, 0 failed, 5 total.

Package	Arch	Status	Sha
sqlite	wasm32	built	5a13e73a
sqlite	wasm64	built	d7ce2dd3
php	wasm32	built	9fd390fa
lamp	wasm32	built	ca2226cc
wordpress	wasm32	built	0eb6e1e7

_{Auto-generated; replaced on each push. Raw data in the publish-status workflow artifact.}

[brandonpayton]

Closing this focused source PR because its unique UTF-16 STAT4 regression work is now carried by #934 at exact green head 260d75d5adc31b327f5591229dbfbb287d395fd6. #934 keeps it as a separate purpose commit and passed the full staging and browser gates. This is not a claim that this branch was merged verbatim.