fix(build): surface clear disk-full error instead of misleading opset ValueError

[timenick]

Closes #259

Problem

When the disk fills during the optimize step of winml build, the ONNX write fails: onnx.save_model truncates the target to 0 bytes and then write() raises OSError(ENOSPC), leaving a corrupt/zero-byte optimized.onnx behind. The quantize step then loads that file — onnx.load parses the empty bytes into an empty ModelProto with no opset_import, and ORT raises the opaque:

ValueError: Failed to find proper ai.onnx domain

which surfaces as Quantization failed: .... The real cause (out of disk space) is never reported, so users chase a phantom code bug.

Fix

Root-cause fix at the ONNX write boundary, plus a defensive guard:

• ONNXSaveError (onnx/persistence.py) — new OSError subclass with path and disk_full attributes. Exported from onnx/__init__.py.
• save_onnx — both write paths (external-data and inline) now catch OSError, remove the partial .onnx/.data artifact, and raise ONNXSaveError. Disk-full (errno.ENOSPC / Windows ERROR_DISK_FULL 112) gets a clear "Insufficient disk space" message; other errors get a generic write-failure message.
• copy_onnx_model (onnx/external_data.py) — same treatment for the copy boundary, cleaning up dst (+ sidecar).
• Quantizer guard (quant/quantizer.py) — _quantize_single_pass now detects an empty/opset-less/unparseable input model up front and returns a clear QuantizeResult failure instead of letting ORT raise its opaque opset error.
• Build hint (commands/build.py) — disk-space errors map to an actionable hint (free up disk / clear ~/.cache/winml).
• Docs (docs/troubleshooting.md) — note on the new behavior and automatic partial-file cleanup.

Subclassing OSError keeps existing except OSError callers working and lets the build command's top-level handler surface the clear message verbatim. The write-boundary cleanup guarantees no truncated artifact is left for a later stage to misread.

Out of scope (decided during triage): no proactive free-disk pre-check before build.