Memory Leak Logs

[payalcha]

Reviewed PR - #1155 (Had to close as some unsigned commits were there)

Openfl component change
Add memory leak logs for aggregator and collaborator, if log_memory_usage flag can be set for ubuntu.

Changes -

1. aggregator.py - Function get_memory_usage added , which collects process as well as system log.
2. collaborator.py - Function get_memory_usage added , which collects process as well as system log.
   Output -
   agg_mem_details.json - details of aggregator in json format
   _mem_details.json - details of aggregator in json format

Structure -
{
"round_number": 0,
"metric_origin": "collaborator1",
"process_memory": 1155.19,
"virtual_memory": {
},
"swap_memory": {
}
}

Test changes

1. log_memory_usage flag is added in conftest.py and passed to test when set.
2. Correction -
   federation_helper.py - _verify_completion_for_participant - added timeout on the basis of num_rounds.

[MasterSkepticista]

@payalcha Per our last discussion, I'm not in favor of this PR for the following reasons:

1. There is evidence of memory growth from internal experiments, which means we need to identify the source of the growth/leak to disposition it. If the investigation suggests changes to openfl to fix the issue, that is a great PR to have.
2. In a situation where the leak/growth is due to an underlying framework (pandas, tf, torch etc.), or it falls outside of openfl purview, it needs to be tracked in the respective framework repository.

That said, if the overall decision is to still add this telemetry functionality, I suggest you use decorators over functions that you want to log memory for, within your CI tests. All helper functions can live within tests/. I don't think framework source code (anything within openfl/) is the right place for it, because users are neither required nor encouraged to use it in their usual experiments.