fix(table): populate caller records on multi-record Insert (#24)

[VojtechVitek]

Related to #24.

TL;DR

Issue #24's original symptom — SQLSTATE 42601 (VALUES lists of differing length) on a batch insert with mixed ,omitempty fields — is already fixed on master by #54 (union columns + DEFAULT). Investigating it surfaced a second, adjacent bug in the same batch-insert path, which this PR fixes.

The bug

Table.Insert with multiple records did not populate the caller's records with DB-generated ids/defaults when the records were passed as individual variadic args:

db.Accounts.Insert(ctx, a, b)   // a.ID, b.ID stay 0  ❌
db.Accounts.Insert(ctx, recs...) // recs[i].ID populated ✔ (spread slice)

insertAll relied on GetAll rewriting the slice backing array. GetAll (scany) resets the slice and appends freshly-allocated structs, so the DB values reach the caller only when they spread a slice — the backing array is shared. Passing individual records builds a fresh internal slice, so the caller's named pointers never see the ids. This is inconsistent with:

• single Insert (insertOne → GetOne scans into the record), and
• Save (saveAll already copies DB values back into the originals).

The fix

Preserve the original pointers and copy the RETURNING * values back into them after GetAll, exactly as saveAll does. Now both calling styles populate ids/defaults.

Commits (red → green)

1. test(table) — end-to-end Postgres regression reproducing omitempty bug in InsertRecords #24 (mixed ,omitempty batch). The 42601 assertion already passes; the populate-back assertions fail. → CI red.
2. fix(table) — the copy-back. → CI green.

This is a real bug fix, so CI is intentionally red on commit 1 (proving the defect) and green on commit 2.

Testing

• New TestInsertMultiPopulatesRecords_Issue24: red before, green after.
• Full suite incl. Postgres-backed ./tests/... passes.
• go build ./..., go vet ./ clean.

Once merged, #24 can be closed (both halves addressed: 42601 by #54, populate-back here).

🤖 Generated with Claude Code

[VojtechVitek]

This was failing with:

=== RUN   TestInsertMultiPopulatesRecords_Issue24
    issue24_repro_test.go:34: 
        	Error Trace:	/home/runner/work/pgkit/pgkit/tests/issue24_repro_test.go:34
        	Error:      	Should not be zero, but was 0
        	Test:       	TestInsertMultiPopulatesRecords_Issue24
        	Messages:   	id populated on caller record