[OpenMP] a newly added loop collapse test fails the correctness check

[vadikp-intel]

openmp/runtime/test/worksharing/for/collapse_many_int.c currently fails (has been marked for expected failure)

the check tests to see if OpenMP executed iterations of the following loop match its scalar execution
for (i = iLB; i <= iUB; i += iStep)
for (j = i * jA1 + jA0; j <= i * jB1 + jB0; j += jStep)
for (k = j * kA1 + kA0; k <= j * kB1 + kB0; k += kStep)

when built with -DVERBOSE and run it will produce the following output:
| Trying iLB=-2; iUB=3; jA0=-7; jA1=-1; jB0=5; jB1=0; kA0=-13; kA1=-2; kB0=37; kB1=-1; iStep=1; jStep=4; kStep=2;
| OpenMP FAILURE: Openmp processed fewer iterations: 540 vs 542
| OpenMP FAILURE: (0 5 31) not processed
| OpenMP FAILURE: (1 4 33) not processed

apparently, for the case above OpenMP may be missing iterations: i=0; j=5; k=31 and i=1; j=4; k=33

[llvmbot]

@llvm/issue-subscribers-openmp

Author: Vadim Paretsky (vadikp-intel)

openmp/runtime/test/worksharing/for/collapse_many_int.c currently fails (has been marked for expected failure)

the check tests to see if OpenMP executed iterations of the following loop match its scalar execution
for (i = iLB; i <= iUB; i += iStep)
for (j = i * jA1 + jA0; j <= i * jB1 + jB0; j += jStep)
for (k = j * kA1 + kA0; k <= j * kB1 + kB0; k += kStep)

when built with -DVERBOSE and run it will produce the following output:
| Trying iLB=-2; iUB=3; jA0=-7; jA1=-1; jB0=5; jB1=0; kA0=-13; kA1=-2; kB0=37; kB1=-1; iStep=1; jStep=4; kStep=2;
| OpenMP FAILURE: Openmp processed fewer iterations: 540 vs 542
| OpenMP FAILURE: (0 5 31) not processed
| OpenMP FAILURE: (1 4 33) not processed

apparently, for the case above OpenMP may be missing iterations: i=0; j=5; k=31 and i=1; j=4; k=33