Minor tweaks for BH cache invalidate

[pgkeller]

Ticket

None

Problem description

No need to invalidate the cache inside a spin loop reading registers.
However, still worth keeping the invalidate after a read barrier as the cost is low and not having a barrier there leaves that to the client code which is error prone and hard to debug.

What's changed

Moved the invalidate outside the spin loops in read barrier paths

Checklist

• Post commit CI passes
• Blackhole Post commit (if applicable)
• Model regression CI testing passes (if applicable)
• Device performance regression CI testing passes (if applicable)
• (For models and ops writers) Full new models tests passes
• New/Existing tests provide coverage for changes

[tt-aho]

We should also update noc_async_full_barrier with this