Cast LLMEncoder output to torch.float32, freeze final layer at init.

[jeffkinnison]

LLMEncoder fine-tuning is currently blocked by two issues:

• RuntimeError: mat1 and mat2 must have the same dtype, but got Half and Float.
• As a result of Add custom prepare_for_training logic to ECD model for LLM encoder adapter initialization #3874, the final layer of the base model is no longer frozen until training time, potentially leading to an "unused parameters" error

This PR adds fixes for both issues:

• LLMEncoder output is now explicitly cast to torch.float32
• The final layer of the base model is always frozen in LLMEncoder.__init__

[github-actions]

Unit Test Results

  6 files  ±0    6 suites  ±0   14m 27s ⏱️ +16s
12 tests ±0    9 ✔️ ±0    3 💤 ±0  0 ❌ ±0 
60 runs  ±0  42 ✔️ ±0  18 💤 ±0  0 ❌ ±0 

Results for commit 6721dfc. ± Comparison against base commit ea50811.