[SYCL] fix incorrect application of binary AND in tests

[cperkinsintel]

&= with 0 is always 0. For tests that are returning 0 as success, using binary AND will mask errors, leading the test to incorrectly pass when it should fail. I wasted an hour last week being mislead by this problem. I audited all our tests and fortunately the ones making this mistake were relatively few.

[aelovikov-intel]

IMO, a better fix is to use += if it returns number of mismatches found, or change the function to return true/false for success/failure.

[cperkinsintel]

I didn't mean to change that to an int, have restored it back to it's original boolean declaration.

I'm disinclined to make any further change. The variable is a boolean named MismatchFound. It's pretty clear what it is. If I were change it to be something that could default to true, it'd have to be MismatchNotFound and then I'd also have to change the func(...) that is called to similarly reverse its interpretation.
It's fine having a boolean initialized to false, it just means the test has to use |= instead of &= .

[aelovikov-intel]

I'm only approving this because it eliminates the immediate buggy behavior.

IMO, the PR should be addressing this on a slightly bigger scope by changing the return types to bool and renaming all the int Result to bool Success to eliminate any ambiguity.

I also don't think that refusal to address that and relying on other contributors to address that technical debt is an exemplary behavior.

[againull]

CI failure on pvc:
AddressSanitizer/common/kernel-filter.cpp
was reported by Chris in CMPLRLLVM-64936

[againull]

@cperkinsintel sycl/test-e2e/Basic/fpga_tests/fpga_pipes.cpp (which was changed in this PR) failed in post-commit, could you please take a look?

FYI There is an old github issue, but I believe that failure hasn't been seen for a while #14308
Test was reenabled here #14627