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

[shinaoka]

Summary

Add backward compatibility support for sparse-ir 1.x API to help users migrate from older versions.

Changes

Backward Compatibility Features

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

Bug Fixes

• Fixed MatsubaraSampling.fit() bug

Merged Changes from mainline

• pylibsparseir v0.8.0 support (PR Update to pylibsparseir v0.8.0 and add derivative support #69)
• deriv() method support (PR Update to pylibsparseir v0.8.0 and add derivative support #69)
• Version 2.1.0 (PR Bump version to 2.1.0 #70)
• Development Status: Production/Stable

API Changes

Before (sparse-ir 1.x)

# Old API with work_dtype
sve_result = sparse_ir.sve.compute(kernel, eps=1e-5, work_dtype=numpy.float64)

# Old API with kernel parameter
basis = sparse_ir.FiniteTempBasis('F', beta, wmax, eps=1e-5, kernel=kernel)

After (sparse-ir 2.x with compatibility)

# Still works! work_dtype is accepted but ignored
sve_result = sparse_ir.sve.compute(kernel, eps=1e-5, work_dtype=numpy.float64)

# Still works! kernel is accepted but deprecated
basis = sparse_ir.FiniteTempBasis('F', beta, wmax, eps=1e-5, kernel=kernel)

# Recommended new API
sve_result = sparse_ir.sve.compute(kernel, eps=1e-5)
basis = sparse_ir.FiniteTempBasis('F', beta, wmax, eps=1e-5)

Testing

• All 95 tests pass
• No breaking changes to existing code

Motivation

This allows packages like dcorelib that depend on sparse-ir 1.x API to work with sparse-ir 2.1.0 without requiring immediate code changes.