Running on Genoa & Milan

Nobody has claimed this yet.

## Constraints

* Normal Queue services both architectures
* Shared Queue only provides Milan architecture

## Considerations

* Code compilation builds should target Milan
* Code can be built on Milan with Genoa specific targeting included, to run on Genoa
* Do results bit-compare between Milan & Genoa:
* Using Milan craylibs then yes, bit compare is expected
* If we use Genoa specific compile flags, then this is not certain, but possible
* Can't use these craylibs yet, due to crashes (issue links: )
* Does this remove the `-S USE_GENOA=true` ?

## Requirements

1. Run large tests on Milan & Genoa
* Configure libfabric differently for these cases
* Configure task_dict sizing of I/O nodes differently
* Configure I/O Memory
2. Compile with Genoa specific flags on Milan to then run on Genoa
3. Run KGO checks on Milan & Genoa (perhaps 2 nodes or 4 nodes) for the same test case

Start by locating the current Normal Queue and Shared Queue configuration, the USE_GENOA setting, and the large-test and KGO check entry points. Compare the existing libfabric, I/O-node task_dict, and I/O-memory settings for Milan and Genoa. Done means large tests and KGO checks run on both architectures, with Milan builds using Genoa-specific flags where required.

Written by the indexing model from the issue text.