Look into matmul transpose - using enums instead of bools and fusing the work into matmul

[bbradelTT]

At some point transpose parameters were added to ttnn.matmul and ttnn.linear with the hope of fusing the transpose or just using the transposed data.

This issue is to look into:

• using enums instead of bools and
• instead of using transpose, just have the work be done inside of matmul (either by transposing internally or reading in transposed values)

Alternatively

• the parameters could be removed.

[marty1885]

Hi, just want to chime in on the context. I think this feature was added as one of my feature requests (#9709). It is a part of the feature set needed to improve performance of the Metalium backend on GGML as GGML uses pre-transposed weights.