uv rejects flash-attn wheels with local versions

[charliermarsh]

This seems to have caused some problems with the flash-attn wheels.

#pyproject.toml
...

[tool.uv]
environments = ["sys_platform == 'darwin'", "sys_platform == 'linux'"]
constraint-dependencies = ["torch==2.5.1"]

[tool.uv.sources]
flash_attn = [
  { url = "https://github.com/Dao-AILab/flash-attention/releases/download/v2.7.3/flash_attn-2.7.3+cu12torch2.5cxx11abiFalse-cp310-cp310-linux_x86_64.whl", marker = "sys_platform == 'linux' and python_version == '3.10'"},
  { url = "https://github.com/Dao-AILab/flash-attention/releases/download/v2.7.3/flash_attn-2.7.3+cu12torch2.5cxx11abiFalse-cp311-cp311-linux_x86_64.whl", marker = "sys_platform == 'linux' and python_version == '3.11'"},
  { url = "https://github.com/Dao-AILab/flash-attention/releases/download/v2.7.3/flash_attn-2.7.3+cu12torch2.5cxx11abiFalse-cp312-cp312-linux_x86_64.whl", marker = "sys_platform == 'linux' and python_version == '3.12'"},
  { url = "https://github.com/Dao-AILab/flash-attention/releases/download/v2.7.3/flash_attn-2.7.3+cu12torch2.5cxx11abiFalse-cp313-cp313-linux_x86_64.whl", marker = "sys_platform == 'linux' and python_version == '3.13'"}
]

Now causes uv sync to error with

error: Failed to parse `uv.lock`
  Caused by: The entry for package `flash-attn` v2.7.3 has wheel `flash_attn-2.7.3+cu12torch2.5cxx11abifalse-cp310-cp310-linux_x86_64.whl` with inconsistent version: v2.7.3+cu12torch2.5cxx11abifalse

Not sure if there is a better way to configure usage of these wheels.

Originally posted by @kleinhenz in #12235 (comment)