feat: add aws_g6e instances

[clumsy]

A new GPU-enabled AWS instance has been added.

More details can be found in the following docs:

• https://aws.amazon.com/ec2/instance-types/g6e/
• https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/using-eni.html#network-cards

Test plan:

• added new unit test

[clumsy]

@d4l3k could you please have a look? Please let me know if any modification is needed, thanks!

[kiukchung]

@clumsy thanks for adding. LGTM. Could you take a look at the failing Lint? IIRC torchx is integrated with lintrunner so you can lint your changes locally for faster dev velocity. See: https://pypi.org/project/lintrunner/

[clumsy]

Sure, @kiukchung, will address the failures.

[clumsy]

Addressed the linter issues, @kiukchung. I had to bump the pinned pyre-check-nightly version - the currently configured one is no longer available. I picked the oldest available in this PR to address the failure (perhaps we want to remove this constraint altogether?), but let me know if you want me to revert this part of the PR.

ERROR: Could not find a version that satisfies the requirement pyre-check-nightly==0.0.101709640814 (from versions: 0.0.101720869228, ...

[kiukchung]

thanks. LGTM

[kiukchung]

@andywag , @yikaiMeta could one of you be so kind to take a look at the pyre and docs build failures and let @clumsy know what the next steps are?

[clumsy]

@andywag, @yikaiMeta Do you have any suggestions? Is the problem fixed so that rebasing fixes the issue? Thanks!

[kiukchung]

Per HUD:torchx pyre, docs-build and python unittests-3.8 are already broken. I manually validated that:

1. pyre check passes on the files changed by this PR
2. doc-build failure is unrelated to this PR
3. python -m unittest torchx/specs/test/test_named_resources_aws.py passes

Merging this PR.