Specify batch size for MME

**What did you find confusing? Please describe.**
How do you specify batch size for MME models?

**Describe how documentation can be improved**
[This blog](https://aws.amazon.com/blogs/machine-learning/optimize-your-inference-jobs-using-dynamic-batch-inference-with-torchserve-on-amazon-sagemaker/) describes using env vars to set batch size and other parameters for a single-model endpoint, however, I haven't found any documentation on setting batch size for individual models within a MME.

**Additional context**
Each model in my MME has a `MAR-INF/MANIFEST.json` within its `model.tar.gz`, so I tried to specify `batchSize` in these files, but I don't think it's being applied.

Start with the linked dynamic-batching blog and the model.tar.gz contents, especially MAR-INF/MANIFEST.json. Determine whether batchSize is supported per model in a multi-model endpoint and document the correct configuration or limitation, including what users should expect when the manifest setting is ignored.

Written by the indexing model from the issue text.