Refactor `xnn_tensor_get_size` into a helper function `xnn_datatype_get_size_bits`. Also makes things a bit more generic (if we add 2 bit datatypes, it should just work, assuming they align the same way that 4 bit types do).

[copybara-service[bot]]

Refactor xnn_tensor_get_size into a helper function xnn_datatype_get_size_bits. Also makes things a bit more generic (if we add 2 bit datatypes, it should just work, assuming they align the same way that 4 bit types do).

FUTURE_COPYBARA_INTEGRATE_REVIEW=https://github.com/google/XNNPACK/pull/5974 from mcr229:reshape_softmax 0578f69d6ea218988bc6a6ceda1acf62ba2a6008