If it is smaller than 500 * (Maximum number of threads), the method reverts to the original
unthreaded version of the function. This old function is also the one used if the input
array has random access iterators.

This should fix slow python.api tests and hopefully should get rid of the timeouts on build servers.

Update to clang-format-5 and reformat the codebase 

This uses openmp threads to speed up the `Strings::join()` function which was causing CreateWorkspace to be excessively slow when writing to the algorithm history.
The speedup is not fantastic as it still takes about 8s to create a 1000x10000 workspace (using 8 threads), but it is an improvement on the 60s yielded by the original function.

A new version of the `join()` function was actually added to `Strings.h`, and is used when the iterators passed as arguments are not random access in nature. If this is the case, then the original function is used.

A test was added to `StringsTest.h` to make sure both `join()` functions are being tested.

This is code duplicated in GroupDetectors and ReflectometryReductionOne2.

Re #22315

re #15164

and tackle unsused parameters, and enable to string shortning
re #15164

and make use in history and logging

re #15164

re #15164

The extractToEOL Kernal namespace method has been moved to the Kernal::Strings namespace where it makes more sense to be alongside other string and ascii stream methods

The change was made because of the difficulties i was having with the binary files stream provided by the FileDescriptor.

This is quite a useful one, as I've seen at least 3 different implementations
of that. Rigorous tests for it were added as well.

Refs #7772

Before making real code changes.