Add pure tests for more operators with Java module overrides

[lemmy]

Check the Java module overrides of 22 more operators exhaustively against pure TLA+ definitions over small inputs, as was done for SelectInSubSeq/SelectLastInSubSeq in #132:

• SequencesExt: SelectInSeq, SelectLastInSeq, Contains, IsPrefix, Suffixes, RemoveFirst, RemoveFirstMatch, FoldSeq, FoldLeft, FoldRight, FoldLeftDomain, FoldRightDomain, SetToSeq
• FiniteSetsExt FoldSet, Functions FoldFunction/FoldFunctionOnSet, BagsExt FoldBag, Bitwise shiftR, DyadicRationals Reduce (via Add and Half, since Reduce is LOCAL), IOUtils atoi, SVG PointOnLine, VectorClocks CausalOrder

Folds with an unspecified order are only compared on operators for which the order does not matter. SetToSeq and CausalOrder are defined via CHOOSE, so their results are checked for membership in the set they choose from.

Not covered: operators whose TLA+ definitions are placeholders (TRUE or an arbitrary CHOOSE), i.e. most of IOUtils, CSV, GraphViz, Statistics, and SVG, and the Json operators, which TLC's own tla2tools.jar overrides.

The tests sidestep two disagreements between overrides and their definitions:

• PointOnLine: the override truncates towards zero whereas \div rounds down, so the results differ for negative coordinates, e.g., x = 0 to x = -1 with segment = 2 gives 0 vs. -1. The test is restricted to non-negative coordinates.
• CausalOrder: the override fails with Cannot cast FcnRcdValue to TupleValue on a sequence represented as a function, e.g., [i \in 1..4 |-> ...]. The test passes tuple literals.