Add requests and limits to grading pods launched

[CatarinaSilva]

Description

We should add requests and limits to the grading pods launched, to avoid launching pods when no resources are available and exhausting the cluster resources.

Suggestions:

Good - add requests and limits based on cluster node sizes and average requirements of units
Better - set unit requests and limits and pull from these so we can easily manage units like BLU09
Perfect - set up an autoscaling group that can scale nodes so that when we have a lot of queued grading we can flush it

Current status

Pods are launched from this line and the requests and limits are commented out:

portal/portal/grading/services.py

Line 84 in a84805f

# "--requests='cpu=100m,memory=512Mi'",

[Gustavo-SF]

#255 and #256 for reference