ofRectangleTests: extended == / != coverage (follow-up to #8567)

[danoli3]

Follow-up to #8567 (now merged: operator!= returns !(*this == rect)). Rebased on master, so this PR is just one commit extending tests/types/ofRectangleTests.

Added test coverage

Every case checks that operator!= is the negation of operator==, and where the answer is well defined, that operator== gives it:

• each field differing on its own; identical negative floats
• 1 ULP at different magnitudes/signs: 1, 100, 1920, -3.5, FLT_MAX; 0.1f + 0.2f vs 0.3f; 4 ULP and 0.1 vs 0.1000001 stay unequal
• integer coordinates: 0, ±1, 7, 255, 1080, 1920, 3840, 65535, ±1e6 (vs itself and vs +1), INT_MAX, INT_MIN, and 2^24 where neighbouring ints map to the same float / 1 ULP apart
• +0 vs -0, 0 vs denorm_min, FLT_MIN
• non-finite values (±inf, NaN): only operator agreement is required, since ofIsFloatEqual is not meaningful there (inf - inf is NaN)
• scale / translate / scaleFromCenter round trips
• 200,000 seeded random pairs (exact copies, 1-3 ULP nudges, random floats in ±1e6, random ints), compared in both directions

Verification

Compiled this test main.cpp (with a thin ofxUnitTests shim) against the real ofRectangle.cpp from before #8567 and with the fix:

• before Fix ofRectangle::operator!= disagreeing with operator== #8567: 89 passed, 11 failed (all on the complement property, incl. 17,033 random-pair disagreements)
• with the fix (current master): 100 passed, 0 failed

A larger standalone sweep (4M comparisons) showed 166,430 ==/!= contradictions before the fix and 0 with the fix. The full test project builds/runs in CI via scripts/ci/linux64/run_tests.sh.

Behaviour notes (inherited from ofIsFloatEqual, not introduced here)

• A rectangle with an infinite field is not == to itself, so r != r is now true for it (was false); +inf == -inf is true.
• The tolerance is eps * |a| (~1 ULP), so drift of a few ULP (e.g. 10 × translate(0.1) then -1) still compares unequal.