[wallet] Cleanup tests #7193

This is a follow up of #4906.

Please review commit-by-commit:

• First commit: The test case is supposed to verify that input pruning works. However, for inputs [100] * 2, [10] * 12 and a target value of 221 the old code as well as the new code always return [100] * 2, [10] * 3, regardless of the number of iterations, etc. To actually test pruning, one needs a sufficient amount of "big coins" so that the random approximation misses at least one big coin even if iterations=1000 (default). Thus, "smaller coins" get added by the algorithm. (But the algorithm does not yet know they are not sufficient and will later be useless, after executing the second pass)
• <strike>Second commit: Use clang-format because the code style is really messed up here. (Reminder to use ?w=0 flag on GitHub to disable white space diff) // moved to a later pull

If someone feels like testing that pruning the trimming test must fail on 0.12, git diff 08000~1..fafd093 src/wallet/test/wallet_tests.cpp | git apply onto 0.12 may come in handy.

Confirmed as move-only expect of the new pruning_in_ApproximateBestSet test (L339-L346). utACK.

utACK Those changes make sense to me.

Backported to 0.12 as bfdaa3c - 96e8d12 is also in there, so the test doesn't fail on 0.12.