[iwyu] Take load average into consideration on CI

See merge request !151

[IO] fix IoCompleterBehavior::none

See merge request !161

[gitlab-ci] Fix test-with-stats

See merge request !159

[io/Future] take a const reference of the Future::Callback

See merge request !163

Currently we always resubmit a possible partial completed Future in
the globalIo context.
This is not possible any more when we don't have a running completer.
Therefore we must resubmit on a worker's IoContext.

Furthermore we now select the callerEnvironment at Runtime because
I can not remember the reason why I choose to always resubmit
on the GlobalIoContext.

The EMPER option Meson option for stats is called 'stats' not
'worker_stats'.

This required that we backport iwyu_tool.py from
https://github.com/include-what-you-use/include-what-you-use/pull/891
into tools/, which supports --load.

Only create, initiateTermination and waitUntilFinished if the
completer thread is actually used.

Fix race condition in io::Stats

See merge request !160

The workerStats std::vector was modified concurrently without being
synchronized. The Stats are constructed together with IoContext and
the worker IoContexts are constructed concurrently at the beginning of
the worker loop. And in the Stats constructor the workerStats
std::vector was modified by adding the Stats instance that is
currently being constructed to it.

Turns out, we don't need the workerStats data structure at all. We
simply provide the global and worker IoContexts to
io::Stats::printStats(). This results IMHO in a cleaner design of
printStats() since there is no longer a data structure called
workerStats, that in fact, not only holds the worker's stats. And
while we at it, we rename io::Stats::printWorkerStats() to
io::Stats::printsStats(), since it does, in fact, not only print the
worker stats, but more or less all related worker stats.

[EchoClient] collect the total amount of unexpected echos

See merge request !157

My first test runs don't show unexpected echos. Therefore I think
unexpected echos are rare and their average will always be 0 even if
some sparse unexpected echos occur.

[EchoClient] changes to easier debug our IO evaluation artifact

See merge request !156

* Use emper::lib::math::RunningAverage instead of doing it manually
* Count the unexpected Echos, print a error to the log and go continue
* Return a duration and error message pair from our clientIteration* functions
* Print the current iteration on error
* Wait 10 seconds after an error before terminating

This seams more intuitive. A macro called EMPER_BUILD_STR actually returning
a string and not a lambda which when called returns the built string.

[EchoClient] improve unexpected echo output

See merge request !155

Build up output string and don't incrementally output it to stderr
to hopefully prevent error messages interleaving.
Only output the least significant byte of the unsigned int cast.
Always use two digits per byte and remove the whitespace.
Should have used printf in the first place.

[EchoClient] output buffers if they do not match

See merge request !154

[EchoClient] output the number of all echos done

See merge request !153

[tests] restructure tests directory

See merge request !152

Move io and lib related tests into io/ and lib/.

Our test dictionaries not must have a 'source' key with an array
created with files().
This ensures we use always correct paths regardless of the actual directory
we are currently in.

Until https://github.com/mesonbuild/meson/issues/8608 is not solved
we have to separately specify the tests name in a 'name' key.

[Debug.hpp] wrap actual logging in log-level constexpr

See merge request !150

According to godbolt.org

do { if constexpr(false) { break; } int foo = 42; foo++; } while(false);

does result in code generation for gcc 10.2 and clang 11.0.1 as opposed to

do { if constexpr(false) { int foo = 42; foo++; }} while(false);

which does not result in code generation for both gcc and clang.

And this simple change did indeed significantly increases our echo
benchmark performance. We were probably creating a lot of std::stringstream
objects without ever using them.

[Makefile] Make ninja consider the load average

See merge request !149

[Runtime] print stats in the destructor not at exit

See merge request !147

Various test-runner improvements

Closes #14

See merge request !142

[io/Stats] introduce named constants for special workerIds

See merge request !148

Allocate the Actor and the BPS on the heap as otherwise Fibers may
access already destructed Actor data.

While this does not fix anything, it appears sensible that we call
stop() prior destruction.

This should, at least for the EXIT_SUCCESS case, fix #14.