Allow setting the shared memory (shm) size for TaskAPI

[sevro]

Description

Allow setting shared memory for TaskAPI similar to other inference APIs.

Motivation

Currently it is not possible to use more than one PyTorch dataloader to train a model because of the shared memory limits, they fail with the error:

ERROR: Unexpected bus error encountered in worker. This might be caused by insufficient shared memory (shm).

Additional context

This was previously discussed in "Allow setting the shared memory (shm) size for the API container" #1638.

[miguelvr]

this one should be easy, thanks for reporting!