Do not cache attention requests count

What does this MR do and why?

In #357385 (closed), it was reported that the cached count is inaccurate with the actual results.

In !84140 (merged), the count query has been improved (from ~400ms down to ~14ms).

Since the query is performing better, we don't need to cache the count.

This is behind uncached_mr_attention_requests_count feature flag. The invalidation calls are not behind the feature flag to keep it simple. When we decide to remove the feature flag, we can also remove the calls for User#invalidate_attention_requested_count.

MR acceptance checklist

This checklist encourages us to confirm any changes have been analyzed to reduce risks in quality, performance, reliability, security, and maintainability.

• I have evaluated the MR acceptance checklist for this MR.