test_date_custom_format_supports_nanos_with_length fails on macOS (aarch64-darwin)

Nobody has claimed this yet.

### Environment
- lsd version: 1.2.0
- OS: macOS (aarch64-darwin / Apple Silicon)

### Description

The integration test `test_date_custom_format_supports_nanos_with_length` fails when building lsd on macOS. The test creates two files and expects their timestamps to differ at the nanosecond level, but APFS does not expose sub-millisecond timestamp precision to userspace — both files end up with timestamps that are identical at that resolution, so the assertion fails.

```
failures:
test_date_custom_format_supports_nanos_with_length

test result: FAILED. 42 passed; 1 failed; 0 ignored
```

### Expected behavior

The test should either be skipped on platforms that don't provide nanosecond-precision timestamps (macOS, BSD), or rewritten to not rely on filesystem nanosecond resolution.

### Suggested fix

Guard the test with `#[cfg(not(target_os = "macos"))]`, or use a synthetic timestamp instead of relying on `mtime` nanoseconds from `std::fs`.

Start by locating the integration test `test_date_custom_format_supports_nanos_with_length` and run it on macOS to reproduce the timestamp-resolution failure. Review how it creates files and reads their mtime values, then either guard the test for unsupported platforms or replace the filesystem-dependent timestamp with a synthetic value. Done means the test passes on supported platforms without failing on macOS or BSD.

Written by the indexing model from the issue text.