Refactor: Performance improvements

[ClausHolbechArista]

Change Summary

Performance improvements in eos_designs and common schema tooling

Proposed changes

Based on cProfiling this focuses on the most time impacting areas

• Avoid natural sorting internal data structured when not needed.
• Move schema build to Python instead of Ansible.
• Compile a fully resolved schema and store as "pickle" when building schemas.
• Load schema store from fully-resolved pickles instead of YAML with $ref.
• Adjust schema tooling to not run through resolving schemas during validation

How to test

• No changes to molecule outputs (so conversion works as expected)
• Runtime is better (390 devices, 12 forks):

  arista.avd.eos_cli_config_gen : Generate eos intended configuration---------------- 488.45s -> 473.17s
  arista.avd.eos_designs : Generate device configuration in structured format ------- 558.30s -> 149.62s
  arista.avd.eos_designs : Set eos_designs facts ------------------------------------ 199.58s -> 46.85s

Checklist

User Checklist

• N/A

Repository Checklist

• My code has been rebased from devel before I start
• I have read the CONTRIBUTING document.
• My change requires a change to the documentation and documentation have been updated accordingly.
• I have updated molecule CI testing accordingly. (check the box if not applicable)

[gmuloc]

LGTM once CI finishes
⚡ ⚡ ⚡

[carlbuchmann]

LGTM, once CI completes!