Fix exceeded() to fall back to configured interval when Retry-After is missing

[evan-burrell]

Summary

When Limit::exceeded() is called without a releaseInSeconds value (e.g. when a 429 response has no Retry-After header), the expiryTimestamp was not set explicitly. This caused it to be lazily recalculated from "now" on each call to getExpiryTimestamp(), which can drift and produce inconsistent state depending on when it's accessed.

This change makes exceeded() fall back to the limit's configured releaseInSeconds interval when no explicit value is provided, ensuring the expiry timestamp is always set deterministically at the moment the limit is exceeded.

Before

// Retry-After header missing → releaseInSeconds is null
$limit->exceeded(releaseInSeconds: null);

// expiryTimestamp is NOT set — lazy calculation from "now" on each access
$limit->getExpiryTimestamp(); // recalculates each time

After

// Retry-After header missing → releaseInSeconds is null
$limit->exceeded(releaseInSeconds: null);

// Falls back to the limit's configured interval (e.g. 60s for custom limiters)
// expiryTimestamp is set once, deterministically
$limit->getExpiryTimestamp(); // stable value

Changes

• src/Limit.php: In exceeded(), fall back to $this->releaseInSeconds when $releaseInSeconds is null
• tests/Unit/LimitTest.php: Added 3 tests covering the fallback behavior for regular limits, explicit values, and custom limiters
• tests/Pest.php: Added toLookLike custom Pest expectation that tolerates ±1 second timestamp drift
• tests/Feature/HasRateLimitsTest.php: Replaced exact timestamp assertions with toLookLike to fix flaky CI failures on slow runners

Test plan

• All existing tests pass (84 passed, 0 failures)
• New tests verify fallback to configured interval when releaseInSeconds is null
• New tests verify explicit releaseInSeconds still takes precedence
• New tests verify Limit::custom() defaults to 60s fallback
• Flaky timestamp assertion on Windows CI fixed with tolerant toLookLike expectation

[JonPurvis]

👍 Thanks for this @evan-burrell

[Sammyjo20]

Thank you!

[evan-burrell]

@Sammyjo20 Just added an extended expectation for clock based tests that might be flaky

[Sammyjo20]

Thank you!