fix: tolerate unencodable default_scalar in ShardingCodec.validate

[inchang-ing]

Fixes #4511

Root cause

When creating an array, codec chains are validated by threading an ArraySpec through each codec's resolve_metadata:

• ArrayV3Metadata.__init__ threads the array's fill value (which defaults to dtype.default_scalar()) through the codec chain,
• ShardingCodec._validate_inner_codecs threads the inner codec chain with dtype.default_scalar() as the probe fill value.

Value-transforming codecs (ScaleOffset, CastValue) encode the fill value as part of that threading. When the chain maps only a subrange of the dtype — e.g. float32 values known to lie in [100, 200], scaled to [0, 65535] and cast to uint16 — the probe value (0) scales to -65535, which CastValue rejects:

ValueError: Value -65535 out of range [0, 65535]. Set out_of_range='clamp' or out_of_range='wrap' to handle this.

Array creation failed even though no real value ever needs that encoding: the failure is a property of the probe fill value, not of the codec configuration.

Fix

Make the fill-value threading best-effort in ScaleOffset.resolve_metadata and CastValue.resolve_metadata: if encoding the fill value raises a value-domain error (ValueError/OverflowError), carry the fill value forward unchanged instead of failing. Shape/dtype threading and per-codec validate() are untouched, so:

• structurally invalid configurations still raise at creation time (wrong dtype for a codec, scale=0, out_of_range='wrap' with a float target, ...) — verified below;
• genuine value-domain errors for actual chunk data still surface in _encode_sync at write time (writing a value outside [100, 200] with this chain still errors, as before).

The same tolerance also fixes the runtime pipeline-evolution path, which threaded the same unencodable fill through resolve_metadata.

Changes

• src/zarr/codecs/scale_offset.py: tolerate value-domain failures when encoding the fill value in resolve_metadata
• src/zarr/codecs/cast_value.py: same
• tests/test_codecs/test_scale_offset.py: creation succeeds with an unencodable default fill value (uint16 + ScaleOffset(offset=100)); invalid config still raises
• tests/test_codecs/test_cast_value.py: the issue's exact scenario (float32 [100, 200] → uint16 via ScaleOffset+CastValue) creates and round-trips; invalid config still raises
• tests/test_codecs/test_sharding.py: sharded array creation succeeds when the inner codec chain cannot encode default_scalar

Verification

• Reproduced the issue on main (v3.4.2.dev14): the issue's repro script raised ValueError: Value -65535 out of range [0, 65535].
• With the fix: the repro creates successfully; a full-chunk write and round-trip works; a sharded array with the same inner chain creates and round-trips.
• Regression discipline: reverting the fix makes the 3 new creation tests fail; re-applying makes them pass.
• pytest tests/test_codecs/ tests/test_metadata/test_v3.py: 757 passed, 1 skipped, 2 xfailed.
• pytest tests/test_array.py tests/test_codec_pipeline.py tests/test_unified_chunk_grid.py: 2737 passed, 756 skipped.
• Structurally invalid configs still rejected: ScaleOffset on a string dtype, out_of_range='wrap' with a float target, and scale=0 all raise ValueError at creation.
• A towncrier changelog fragment will be added with the PR number once known.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 94.71%. Comparing base (ec2c359) to head (d6a2e71).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #4526   +/-   ##
=======================================
  Coverage   94.71%   94.71%           
=======================================
  Files          94       94           
  Lines       13663    13669    +6     
=======================================
+ Hits        12941    12947    +6     
  Misses        722      722           

Files with missing lines	Coverage Δ
src/zarr/codecs/cast_value.py	98.75% <100.00%> (+0.02%)	⬆️
src/zarr/codecs/scale_offset.py	100.00% <100.00%> (ø)

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.