The fix leaves an open question: should we keep treating refs
to static array of chars as strings, or should we instead
use `strnlen` to check if it is null-terminated within the buffer

Fixes #1238

Also adds `Predicate` helper function to create `PredicateMatcher`.
Because of limitations in type inference it needs to be explicitly
typed, like so
`Predicate<std::string>([](std::string const& str) { ... })`.
It also takes an optional second argument for description of the
predicate.

It is possible to infer the argument with sufficient TMP, see
https://stackoverflow.com/questions/43560492/how-to-extract-lambdas-return-type-and-variadic-parameters-pack-back-from-gener/43561563#43561563
but I don't think that the magic is worth introducing ATM.

Closes #1236

Specific platforms (e.g. TDM-GCC) can have terrible timer resolution,
and our checking code will then loop for an inordinate amount of time.
This change will make it so that the calibration gives up after 3
seconds and just uses the already measured values.

This leaves one open question, how to signal that the resolution
is terrible and benchmarking should not happen?

Fixes #1237

There are still some holes, e.g. we leave surrogate pairs be
even though they are not a part of valid UTF-8, but this might
be for the better -- WTF-8 does support surrogate pairs inside
text.

Closes #1207