[package update] upgrade awscrt to 0.31.1 or greater to fix aws CLI.

[eap]

awscrt needs to be upgraded to 0.31.1 or greater. Certain AWS calls fail with the following error

+ aws s3 cp /tmp/build_logs.html s3://jedi-ci-build-public-logs/oops-3172-b3291bd-intel-PVM5JWCH.html --content-type text/html
--
upload failed: ../tmp/build_logs.html to s3://jedi-ci-build-public-logs/oops-3172-b3291bd-intel-PVM5JWCH.html 
An HTTP Client raised an unhandled exception: module 'awscrt.checksums' has no attribute 'crc64nvme'

This appears to be a somewhat new issue in the AWS CLI as documented here and is caused by the AWS CLI failing to properly specify its internal package-version compatibility. We fixed this in jedi-ci by manually overriding the package version (see JCSDA-internal/jedi-ci#50) but the underlying issue continues to affect the spack-stack version of aws cli depending on the specific site and which version of awscrt was used.

Long term fix here is to update aws crt. The exact version is unclear due to the bug in boto, but 0.31.1 is a known good version.

[climbfuji]

Does this need to be fixed in spack-stack-2.1? (We started rolling out on the HPCs.) Or sufficient for develop/spack-stack-2.2.x?

[climbfuji]

I assigned this to spack-stack-2.2.0, because there's no time to fix it for spack-stack-2.1.0

[eap]

I'll be applying a manual fix in our container. Its a bug we can work around. 2.2 is fine.

[eap]

Temporarily this has been fixed in the container by downgrading aws tools. I've also added a more recent version of awscrt upstream in spack-packages to allow this to be fixed in the future. The concretizer should handle this correctly once we get py-setuptools updated.