Prepare SQLite Tcl all audit artifacts

[brandonpayton]

Summary

Prepare the SQLite public Tcl all target for the next high-throughput Kandelo audit.

This PR does not claim the 10,523-job all run has passed. It sets up the reviewable harness/reporting pieces needed before that run starts:

• Adds durable per-host outcome lists for SQLite official Tcl runner reports:
  • outcome-lists/passed-jobs.tsv
  • outcome-lists/failed-jobs.tsv
  • outcome-lists/skipped-jobs.tsv
  • outcome-lists/incomplete-jobs.tsv
• Links those outcome lists from the combined project-unit summary.
• Documents the SQLite Tcl all run plan and updates the porting guide.

Target Scope

Target: SQLite public Tcl test/testrunner.tcl all on Kandelo.

all is distinct from Tcl full: full runs the public Tcl scripts once under the default configuration, while all includes full plus SQLite's public runtime-configuration matrix from testrunner_data.tcl.

Current public Tcl job-count reference from the SQLite taxonomy audit:

Target	Public Tcl jobs
full	1416
all	10523

The next runner should still run --explain against a fresh packages/registry/sqlite/sqlite-full-src/ tree and report the actual scheduled testrunner.db count before the long run.

Host Plan

Node.js and browser are separate Kandelo product surfaces. The planned all run should keep them in separate result roots so host-specific failures, Vite/browser restarts, and testrunner.db state cannot be conflated.

The run-plan doc records the intended shape:

scripts/run-sqlite-project-unit-tests.sh --host node --permutation all --explain --results-root <result-root>/explain-node
scripts/run-sqlite-project-unit-tests.sh --host browser --permutation all --explain --results-root <result-root>/explain-browser

scripts/run-sqlite-project-unit-tests.sh --host node --permutation all --jobs <node-jobs> --timeout-ms <timeout-ms> --results-root <result-root>/node-all
SQLITE_TEST_VITE_PORT=<port> scripts/run-sqlite-project-unit-tests.sh --host browser --permutation all --jobs <browser-jobs> --timeout-ms <timeout-ms> --results-root <result-root>/browser-all

Artifact Requirements

Every substantive all run should publish command lines, environment, result roots, summary.txt, combined-summary.md, host-status.tsv, testrunner.db, testrunner.log, failures.tsv, and the four outcome-list TSV files.

Skipped rows are SQLite omit jobs with the exposed reason when present. Incomplete rows are jobs left ready, running, halt, empty, or otherwise nonterminal when the report is written.

Out Of Scope

This PR is for public Tcl testrunner.tcl coverage only. It does not claim coverage for:

• TH3
• SQL Logic Test
• dbsqlfuzz
• release
• mdevtest
• sdevtest

Those are private/proprietary suites, separate public repositories/harnesses, proprietary fuzzing surfaces, or source-rebuild targets rather than Kandelo guest Tcl testrunner.tcl permutations.

Validation

• bash -n scripts/write-sqlite-official-outcome-lists.sh scripts/run-sqlite-official-tests.sh scripts/run-browser-sqlite-official-tests.sh scripts/run-sqlite-project-unit-tests.sh
• git diff --check
• scripts/write-sqlite-official-outcome-lists.sh /tmp/does-not-exist-kd-wdo.db /tmp/kd-wdo-outcome-empty
• scripts/write-sqlite-official-outcome-lists.sh test-runs/sqlite-project-unit-all/kad-29m-explain-20260614-235828/node/testrunner.db /tmp/kd-wdo-outcome-test
  • archived explain DB produced header-only passed/failed/skipped lists and 10,353 incomplete data rows, matching that historical stale explain artifact.
• bash scripts/run-sqlite-project-unit-tests.sh --help
• bash scripts/run-sqlite-official-tests.sh --help
• bash scripts/run-browser-sqlite-official-tests.sh --help
• /Users/brandon/src/kandelo-gascity/tools/kandelo_pr_remote_policy.sh patrol

Not run: the full SQLite Tcl all execution. That is the downstream high-throughput audit run this PR prepares.