Bump version to 2.1.1

[shinaoka]

Summary

Bump version from 2.1.0 to 2.1.1

Changes in 2.1.1

Backward Compatibility (from PR #71)

• work_dtype parameter support in sve.compute(): Accepts (but ignores) this parameter for backward compatibility with sparse-ir 1.x
• kernel parameter support in FiniteTempBasis.__init__(): Accepts kernel parameter (deprecated) for legacy code compatibility
• max_size parameter: Made optional with keyword-only argument

Bug Fixes

• Fixed MatsubaraSampling.fit() bug

Version Rationale

Using 2.1.1 (patch) because:

• Adds backward compatibility without changing existing behavior
• Fixes a bug in MatsubaraSampling.fit()
• No new features, only compatibility improvements

Compatibility

This release allows legacy code (like dcorelib) using sparse-ir 1.x API to work with sparse-ir 2.x without modification:

# Legacy code still works!
sve_result = sparse_ir.sve.compute(kernel, eps=1e-5, work_dtype=numpy.float64)
basis = sparse_ir.FiniteTempBasis('F', beta, wmax, kernel=kernel)

Testing

• All 95 tests pass
• No breaking changes

Related PRs

• Add backward compatibility for sparse-ir 1.x API (work_dtype and kernel parameters) #71: Add backward compatibility for sparse-ir 1.x API