[docker] Fix the nightly docker build

[xuzhao9]

After the submodule update, the FA3 CUTLASS kernels now cost much more memory to build and will exhaust the CI machine.

Make the following changes:

1. Build FA3 with MAX_JOBS=4 and NVCC_THREADS=1.
2. Disable xformers build as it also requires FA3, we will try to enable its build later.
3. Also disable colfax and TK build for now - will fix that later.

Test plan: CI

[facebook-github-bot]

@xuzhao9 has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator.

[facebook-github-bot]

@xuzhao9 merged this pull request in 5cc3976.