fix: make postgres table name configurable via pg_table_name config

[evanh]

There is an idea to have separate tables when using postgres, to isolate some of the work. Make that configurable.

• Adds pg_table_name config field (defaults to inflight_taskactivations for backwards compatibility)
• Migration SQL uses a {TABLE_NAME} placeholder; build_migrator() substitutes the name at runtime and recomputes SHA-384 checksums so sqlx migration tracking works correctly
• All queries in PostgresActivationStore use format! with the configured table name instead of a hardcoded string
• No new dependencies — sha2 was already in Cargo.toml

Test plan

• cargo build passes
• All 200 tests pass (cargo test), including both case_1_sqlite and case_2_postgres variants

🤖 Generated with Claude Code

[evanh]

I don't actually know if this is better than using different databases, which we can already do without any code changes.

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

[markstory]

I don't actually know if this is better than using different databases, which we can already do without any code changes.

What are the downsides of using multiple databases? Are they more complex to provision and maintain?

[evanh]

What are the downsides of using multiple databases? Are they more complex to provision and maintain?

They have different connection pools seems to be the main one. Otherwise the main limitation is that you can't query across DBs, but that is not an issue for us.