Add batchnorm folding transformations (fx) (#348)
* added batchnorm folding
* doc and renaming
* slightly better doc
* handle special linear and batchnorm1d case
* add batchnorm1d before linear
* moved test
* fix
* add parameterized as a test requirement
* use get_submodule
* add test
* better doc
* install tests dependencies for fx tests
* move back transformations tests to subfolder
* remove levit transform that is too custom
* remove levit test as we won't support its custom transform
* fix fx test yml
* remove fx quantization tests as they were not run previously, and are broken (to fix in an other PR)
* remove deepcopies
* readd comment
* style
* refactor test
* remove duplicate arg
* simplify tests