[FEATURE REQUEST] `dot` test coverage and fixtures

[LoserCheems]

Problem statement

The dot kernel currently has no dedicated tests under tests/. While copy and swap are covered by test_copy.py and test_swap.py, there is no test_dot.py to verify that all backends compute $z = x^\top y$ correctly.

Without dot tests:

• regressions in any backend implementation may go unnoticed,
• CI cannot guarantee numerical equivalence between backends and the Python reference,
• the README BLAS table cannot mark the Test column for dot as complete.

Proposed solution

Add a dedicated pytest module for dot that follows the style of existing kernel tests and validates all backends.

Concretely:

• Introduce tests/test_dot.py alongside the existing tests.
• For each backend (Python, PyTorch, Triton, CuTe when available), compare results against the Python dot reference for multiple input sizes and dtypes.
• Cover edge cases such as zero-length vectors, large values, and orthogonal vectors where the dot product is close to zero.

Alternatives considered

Ad-hoc or manual testing is not sufficient because it:

• is easy to skip during development,
• does not integrate with CI,
• lacks consistent regression coverage.

Implementation details

• Add tests/test_dot.py using pytest patterns already used in the project.
• Reuse any shared helpers for generating random vectors and comparing outputs.
• Parameterize tests over vector length, dtype, and value ranges.
• Gracefully skip backend-specific tests when a backend is not available or not yet implemented.

Use case

The dot tests will:

• validate correctness for all backend implementations of dot,
• support contributors in implementing and optimizing dot product kernels,
• reinforce the project’s emphasis on numerical equivalence across backends.

Related work

• Existing tests: tests/test_copy.py, tests/test_swap.py.
• BLAS dot test suites in other numerical libraries.

Additional context

This issue is part of filling in the dot row of the README BLAS table, specifically the Test column.