Fix ConfigEps execution provider ordering in test files
When BaseTester::ConfigEps() is called with multiple EPs, it runs the
first available EP for the given operator. CPU EP should be added last
to ensure other EPs get tested first.
Files fixed:
- matmul_2bits_test.cc: Move CPU EP after WebGPU EP
- skiplayernorm_op_test.cc: Move CPU EP after WebGPU EP
- gather_op_test.cc: Move CPU EP after CUDA EP (4 test cases)
Co-authored-by: fs-eire <7679871+fs-eire@users.noreply.github.com>
