release(v2.6.0): aggregates and grouping

[ikelaiah]

What

v2.6.0 release: aggregates and grouping. The feature work landed in #29; this PR is the release commit.

Release commit

• package.json + package-lock.json → 2.6.0
• docs/releases/v2.6.0.md release notes
• CHANGELOG.md: the Unreleased section becomes the v2.6.0 entry
• README quick start, supported-dialects line, and doc index updated
• docs/USER_GUIDE.md runtime ZIP reference; docs/ACCURACY.md version + snapshot link
• docs/RELEASES.md current release
• docs/metrics-v2.6.0.json written via npm run metrics:write

Shipped in v2.6.0

• Per-column aggregates (COUNT, COUNT DISTINCT, SUM, AVG, MIN, MAX); remaining picks become GROUP BY (omitted when all picks are aggregated); deterministic aliases; aggregate ordering; DISTINCT disabled and explained while grouping; fan-out tip.
• Query files version 2 with backward-compatible migration from version 1; aggregates saved, exported, restored, and pruned with their picks.
• ADR-003 and the aggregates section in docs/QUERY_BUILDER.md.

Verification

• Golden 221/221 (query suite 55 records incl. aggregate engine, v1 migration, storage/prune)
• ERD UI suite 68/68 over HTTP; Firefox smoke 10/10
• typecheck, build, test:file, metrics, package:smoke (procflow-v2.6.0.zip), benchmark all green
• After merge: annotated tag on the release(v2.6.0) commit and GitHub release with the runtime ZIP + SHA256SUMS.txt