[kvlonge]

Why

A batch of small bugs found while upgrading a real project: wrong or missing summary counts, a crash when selecting only a Python task, a missing loader guard under sqb load, a planner/build crash for timezone-aware microbatch cursors fed by DATE inputs, headers cut short by ); inside quoted values, and a table-type downgrade prompt after every Snowflake clone.

Changes

• CHI-419: the migrate_from removal notice names the resolved target ('default' when unnamed).
• CHI-416: the build summary counts source, seed and end-of-build audit outcomes.
• CHI-417: sources declared by selected Python tasks and assets are planned, with the same deferral as SQL reads (no more G000).
• CHI-418: sqb load gives loaders the declared-source map, so hard-coded relation names warn; the load summary counts and lists warnings.
• CHI-410: microbatch counting, capping and batch splitting treat a naive bound as UTC when the other bound is aware; MODEL/TEST/FUNCTION/SCENARIO headers only end at an unquoted );, scanning in linear time.
• CHI-420: clones take the destination model's table type, so a permanent production table clones as transient into a transient dev target.

Verification

• New unit, integration and DuckDB E2E tests for each fix; each fails on the old code.
• make test, make check-ci, make check-e2e-shards and the hygiene scan pass.
• One independent review plus a follow-up pass.
• A Snowflake integration test for the clone type is added (real-warehouse, not run in CI).