ObjectEqualsForPrimitives should not suggest to convert boxed equals on double or float values

[findepi]

https://errorprone.info/bugpattern/ObjectEqualsForPrimitives check is described as "performance" related, so if I follow it, I would not expect the semantics to change.
Currently, ObjectEqualsForPrimitives applies to all primitives and enums.
However, double and float values should be exempted from the check:

double a = Double.NaN;
double b = Double.NaN;

System.out.println(Objects.equals(a, b)); // true; flagged by ObjectEqualsForPrimitives check
System.out.println(a == b); // false

In current master of https://github.com/trinodb/trino, the ObjectEqualsForPrimitives triggers typically in equals() implementations. While most double/float values are unlikely to be NaN, this is not something and equals() implementation should be concerned about.

It's probably worth noting that equals generated for Java record classes also compares NaN values as equal (as if Objects.equals was used)

[ksobolew]

Is Double.compare(a, b) == 0 a valid substitute? No boxing, but perhaps a more expensive operation.

[findepi]

@ksobolew you're right.
So the check is probably valid also for float/double, but it maybe its message could be tailored down to prevent people like me from improving the code incorrectly.

[findepi]

the request here is invalid.
the idea to improve error message logged as new issue -- #4392