feat: object store decides scheduler type

[westonpace]

Summary

• Add ObjectStore::prefers_lite_scheduler() that returns true for file+uring:// stores, so the lite scheduler is used automatically without needing the env var
• Change SchedulerConfig::use_lite_scheduler from bool to Option<bool> — Some(true/false) overrides, None defers to the object store's preference
• LANCE_USE_LITE_SCHEDULER env var still works as an override when explicitly set

Test plan

• cargo check -p lance-io --tests --benches compiles cleanly
• cargo test -p lance-io — all 148 tests pass
• cargo clippy -p lance-io --tests --benches -- -D warnings — no warnings

🤖 Generated with Claude Code

[github-actions]

Review

Clean, focused PR. Two items worth addressing:

P1: LANCE_USE_LITE_SCHEDULER env var doesn't respect falsy values

use_lite_scheduler: if std::env::var("LANCE_USE_LITE_SCHEDULER").is_ok() {
    Some(true)
} else {
    None
},

LANCE_USE_LITE_SCHEDULER=0 or LANCE_USE_LITE_SCHEDULER=false will still enable the lite scheduler since .is_ok() only checks presence, not value. Now that the field is Option<bool>, this is a good opportunity to parse the value properly (e.g., treat "0", "false", "" as Some(false) and "1", "true" as Some(true)), so users can also force-disable the lite scheduler via env var when the object store would otherwise prefer it.

P1: No test coverage for the new fallback logic

The tri-state resolution (Option<bool> → object store preference) and prefers_lite_scheduler() are untested. A unit test that constructs a file+uring:// object store and verifies the scheduler selection falls through to prefers_lite_scheduler() when use_lite_scheduler is None would prevent regressions.

[codecov]

Codecov Report

❌ Patch coverage is 98.18182% with 1 line in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
rust/lance-io/src/scheduler.rs	98.00%	0 Missing and 1 partial ⚠️

📢 Thoughts on this report? Let us know!