dataman: Add client sync perf counter and increase default timeout to 5s

[niklaut]

Solved Problem

Dataman requests occasionally time out during boot. My suspicion is that its a combination of slow SDCard and unfortunate scheduling order during boot when the CPU is very busy. The timeout of 1s may be reached more quickly than expected.

However, it is notoriously difficult to reproduce this issue under controlled and debuggable/traceable conditions. I was only able to catch it once and the dataman: write perf counter "only" recorded most=310ms. However, on unrelated traces I've seen scheduling latencies on the dataman thread in the hundreds of milliseconds, here 271ms:

Solution

This increases the dataman client timeout to 5s and logs the time it took to service the sync requests.
There are already two perf counters dataman: write and dataman: read, however, those are running in the dataman thread, while the client is running on another thread. Thus adding this perf counter will show scheduling latencies between the dataman and its client threads.

Changelog Entry

For release notes:

Bugfix Increase dataman client timeout

Alternatives

Dataman request timeouts are bad anyways, since the data is needed to fulfill the mission. There really isn't any alternative to not getting any data.

Context

See also #22778.