Un-skip f32, f64, and float_exprs spec tests

These tests were skipped because NaN canonicality checks were incorrect.
Now that isCanonicalNaN and isArithmeticNaN are fixed, all three test
suites pass.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: sumleo

scripts/test/shared.py

@@ -410,9 +410,6 @@ def get_tests(test_dir, extensions=[], recursive=False):
     'conversions.wast',  # Promoted NaN should be canonical
     'data.wast',    # Fail to parse data segment offset abbreviation
     'elem.wast',    # Requires modeling empty declarative segments
-    'f32.wast',     # Adding -0 and -nan should give a canonical NaN
-    'f64.wast',     # Adding -0 and -nan should give a canonical NaN
-    'float_exprs.wast',  # Adding 0 and NaN should give canonical NaN
     'float_misc.wast',   # Rounding wrong on f64.sqrt
     'func.wast',    # Duplicate parameter names not properly rejected
     'global.wast',  # Fail to parse table