[train] remove documentation for amp

[matthewdeng]

Why are these changes needed?

There has been little usage of these interfaces. Removing them from the documentation to make it clearer what the recommended APIs are.

As a replacement, users may typically want to define this logic themselves using PyTorch directly or Accelerate.

Related issue number

Checks

• I've signed off every commit(by using the -s flag, i.e., git commit -s) in this PR.
• I've run scripts/format.sh to lint the changes in this PR.
• I've included any doc changes needed for https://docs.ray.io/en/master/.
  • I've added any new APIs to the API Reference. For example, if I added a
    method in Tune, I've added it in doc/source/tune/api/ under the
    corresponding .rst file.
• I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/
• Testing Strategy
  • Unit tests
  • Release tests
  • This PR is not tested :(

[pcmoritz]

Yeah, I have a strong feeling that this is not used any more given when it was introduced 80ae651#diff-b286fb90d7135981ab7eeaa0a62ccb8b14cc32a3b0af5e15e7883c0f27bf7f46 -- it is a wonder that this is still working if it in fact is :D

[woshiyyya]

LGTM. Given that almost all PyTorch-derived libraries(lightning, transformers, accelerate, deepspeed) now support mixed-precision training, we don't need to maintain this out-dated API ourselves.