Added counters of already preprocessed assertions

[Tomaqa]

Only matters in incremental solving when it may happen that some assertions are preprocessed multiple times.

If not trackPartitions(), then we do not use the counters and always preprocess all frame formulas again in the hope that it will benefit from the new assertions.

If trackPartitions(), then we use the counters to ensure that every formula is preprocessed at most once.

Requires #888

[Tomaqa]

As done in #888, I add a performance comparison in QF_LRA with the 2-minute timeout. This time I always exclude unsat cores and only include (non-)incremental.

Comparison before and after applying this PR:
2025-v2.9.2-23-g883d2889_v2.9.2-24-ga646c8af.scatter.pdf
.. confirms that when not tracking partitions, there is no effect.

Comparison before and after applying this PR when trackPartitions() is forced to true:
2025-v2.9.2-23-g883d2889-dirty_v2.9.2-24-ga646c8af-dirty.scatter.pdf
.. shows a beneficial trend, especially in the case of unsat benchmarks.

Comparison of the new version with and without tracking partitions:
2025-v2.9.2-24-ga646c8af_v2.9.2-24-ga646c8af-dirty.scatter.pdf
.. shows that in the case of unsat benchmarks, the overhead of the tracking is quite low.

[Tomaqa]

I also tried the variant to preprocess formulas at most once when not tracking partitions:
2025-v2.9.2-24-ga646c8af_v2.9.2-25-g762b3b41.scatter.pdf
However, the results are not conclusive. For these benchmarks, it is beneficial for unsat cases, but not for sat cases. It may be worth, though, to allow configuring this option and leave it to the user.