Add deterministic scheduler interval regression tests

[samrusani]

Description of Changes

Closes #5591.

• Retain the original scheduled call time across reducer execution and cleanup through a private execution/reschedule seam.
• Add deterministic coverage that simulates completion 30 seconds later and proves the next interval is based on the pre-execution call time.
• Keep one-shot schedules and procedure/error-path rescheduling behavior unchanged, without sleeps or wall-clock tolerances.

API and ABI breaking changes

None.

Expected complexity level and risk

1. This is an internal testability refactor with no public behavior or API change.

Testing

• cargo fmt --all -- --check
• cargo test -p spacetimedb-core reducer_reschedules_from_call_time_not_completion_time
• cargo test -p spacetimedb-core host::scheduler::tests
• cargo test -p spacetimedb-core --lib (151 passed)
• cargo clippy -p spacetimedb-core --all-targets -- -D warnings
• git diff --check
• Temporarily restored completion-time rescheduling; the focused regression test failed as expected, then passed after restoring call-time rescheduling.

[Shubham8287]

interval_reschedule method is too trivial to test. I don't think these tests are really testing anything good.

[samrusani]

Agreed — the helper-only assertion did not cover the execution boundary. I replaced it in 6d01046 with a seam used by the actual scheduled-reducer path: the test simulates completion 30 seconds after the captured call time and verifies that both reschedule timestamps still advance from the call time. As a mutation check, temporarily restoring completion-time rescheduling makes the focused test fail; after restoring call-time rescheduling, the focused test, all 151 core library tests, formatting, and clippy pass.