Bug: is_torch_xla_version() checks function truthiness instead of _torch_xla_available flag

[Ricardo-M-L]

Describe the bug

is_torch_xla_version() in src/diffusers/utils/import_utils.py guards with if not is_torch_xla_available — but is_torch_xla_available is the function defined in the same module, not the cached boolean. A function object is always truthy, so the guard never short-circuits when torch_xla is absent.

When torch_xla is not installed, parse(_torch_xla_version) then operates on an invalid/empty version string instead of returning False.

Expected behavior

The guard should check the module-level boolean _torch_xla_available:

if not _torch_xla_available:
    return False

Found while auditing version checks in import_utils. Fix incoming in PR #14513.