Use boost-ext::ut

[ClausKlein]

see https://boost-ext.github.io/ut/

[dietmarkuehl]

Most of the changes are OK. I'm also not objecting to the direction of using something like boost-ext.ut as a testing implementation: it is certainly an improvement over my ad hoc approach. There are, however, a number of concerns:

1. I would like the tests to be a reasonable basis for tests of standard libraries. When I spoke to library maintainers for libstdc++ they stated that, e.g., depending in tests on <iostream> is a no go for them.
2. Embedding the implementation rather than depending on a separate repository doesn't seem to be the direction Beman generally goes: overall there is discussion on how to support package management, e.g., to bring in a test framework like Google Test for project which wish to do so (for the reason stated above I'm also not keen on Google Test).
3. During today's meeting (2025-04-21) I only mentioned the use of boost-ext.ut very briefly and we didn't have time to get into a discussion on what we imagine for Beman going forward. Next week at C++Now I hope to get more discussion on this topic.

I realise that only the second point is possibly actionable. On the other hand the PR doesn't change the existing tests to move to boost-ext.ut, i.e., it is more an enabling change than something which will make it hard to move into a different direction. I think my current position is that I'd land the change if the dependency were not embedded probably using FetchContent for now which is already used for project_options. Following the previous supply chain attack on github it is probably preferable to pull in the dependencies using a fixed SHA and update when necessary.

[ClausKlein]

First of all, at the moment this MR is a very experimental study only!

see boost-ext/ut#678

I like boost-ext::ut, but I was not expecting so much compile problems.

I will continue to analyse the tests that do not compile yet.