Benchmarks

[inge4pres]

Describe the current state

The current development pace is favoring completeness over performance.
In this regard, some concerns have been raised about the use of locking via mutexes.

Describe the desired state

We should have a benchmark test suite, able to detect regressions in CI but most importantly able to indicate if alternative implementations yield more performant code.

Proposed solution

Ideally we'll have a separate test suite for each signal, at a minimum it should be possible to run the benchmarks for single signal at a time.
The benchmark suite can be structured either with microbenchmarks on the sensitive parts of the signals' API/SDK, or as a end-to-end suite that benchmarks a reference implementation (just like we have example builds).

[inge4pres]

After some research, it seems that benchmarks.zig might be the best choice to run benchmarks.
Ideally, the benchmarks would be executed as tests, but not as part of the same zig build test target, as we want to keep that fast for development.

Proposal: add a zig build benchmark target that will run the benchmarks.
The benchmarks tests will live in their own module and will have to report their execution somehow.

It would be nice to store the results somewhere, in order to have a baseline to compare results with from time to time.
I guess we can keep the benchmarks report as part of released artifacts, so we don't have to worry about it right now.

[inge4pres]

I did a quick spike on the library mentioned above, unfortunately it doesn't build with Zig 0.14.
I started working on integrating zBench, which does support 0.14 and seems able to produce the desired outcomes.