Slow memory leak

[Mark2000]

Describe the bug
The general environment slowly leaks memory.

To Reproduce
Steps to reproduce the behavior:

1. Train the general environment with rllib. After ~5 million steps over 30 cores, the memory usage has increased from about 30% to 100% of 3.8gb/core.

Expected behavior
No leaks.

Version (please complete the following information):

• Python: 3.10
• Basilisk: 2.2.1a?
• Plaform: CURC cluster

[Mark2000]

It appears that this issue is related to ray/rllib and not the environment. At some point, a job should be submitted to CURC that repeatedly resets and steps through the environment without any training code while monitoring memory usage to confirm the environment is not leaking.

[Mark2000]

Still seeing a memory leak just by repeatedly stepping through and resetting env, no training.

[Mark2000]

Using objgraph, a major source of leakiness is reading from bsk loggers/messages. Appears to allocate new vectors for every read that do not get deleted?

[Mark2000]

Switching to #86 may address some memory leaks.