Add RMSNorm (part of parallel_attn_blocks)

[lessw2020]

Summary:
This PR adds RMSNorm.
It is used in parallel_attention_blocks but moving the earlier mono PR for that into smaller chunks for easier review and landing.

Test plan:
added two units tests:
test_rms_norm_fp32
test_rms_core_algo (compares F.norm version vs PR version)

Fixes:
Adds requested doc string to RMSNorm from earlier PR review.

[codecov-commenter]

Codecov Report

Patch coverage: 100.00% and project coverage change: +0.06% 🎉

Comparison is base (951a452) 69.11% compared to head (122f4b9) 69.17%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #448      +/-   ##
==========================================
+ Coverage   69.11%   69.17%   +0.06%     
==========================================
  Files         170      170              
  Lines       11524    11547      +23     
==========================================
+ Hits         7965     7988      +23     
  Misses       3559     3559              

Files Changed	Coverage Δ
tests/modules/layers/test_normalizations.py	100.00% <100.00%> (ø)
torchmultimodal/modules/layers/normalizations.py	100.00% <100.00%> (ø)

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[rohan-varma]

LG! I would just absorb the fp32 type check test into the other one for brevity.

[facebook-github-bot]

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