[Tests] Shorten sycl_iterator and permutation_iterator tests

[danhoeflinger]

Adds infrastructure to be able to shorten some commonly used test infrastructure by

1. allowing scaling of the max_n (100000) and small_n (16)
2. allowing scaling of the step (3.14.15) multiplier

This also corrects a potential error in the search_20 test of sycl_iterator_find which for some value of n, could return the incorrect result due to a unintentional extension of the searched for list by the nature of the input data.

A few tests were scaled in this PR to have fewer iterations with different sizes:

sycl_iterator_find
sycl_iterator_reduce
sycl_iterator_scan
sycl_iterator_sort
permutation_iterator_parallel_merge
permutation_iterator_parallel_stable_sort

Others were just given the opportunity for similar shortening.

These patterns should also be tested via their individual tests, independently from the sycl_iterator type or the permutation_iterator types. Also, the newly added input_data_sweep tests, provide much more coverage for input data processing to dpcpp backends.

[danhoeflinger]

Note: This does not do anything to address time spent in JIT.
Its quite possible that a lot of the time of the test is spent in JIT compilation, and only removing kernel invocations or splitting the test can address this.

[SergeyKopienko]

Should we think about ability to save previous behavior when the macros TEST_LONG_RUN is defined?

[danhoeflinger]

Should we think about ability to save previous behavior when the macros TEST_LONG_RUN is defined?

Yes, perhaps with a we can do a min / max with the scale factor of 1.0 when TEST_LONG_RUN is defined so that we don't shorten a test someone intentionally made longer with the scale factors. Its a good idea.

[danhoeflinger]

Should we think about ability to save previous behavior when the macros TEST_LONG_RUN is defined?

I've now made it so when TEST_LONG_RUN is true, we undo any reduction in test cases from this. We allow scaling factors to still increase the number of tests in this case. None of the current code uses these scaling factors to increase the number of tests, but it is possible in the future, and we wouldn't want to reduce that for TEST_LONG_RUN.

[danhoeflinger]

Actually... it looks like TEST_LONG_RUN=1 is used in the nightly testing, so it kind of defeats the purpose of this PR to turn off the savings in this case if our goal is to speed up the full suite testing.

[danhoeflinger]

Actually... it looks like TEST_LONG_RUN=1 is used in the nightly testing, so it kind of defeats the purpose of this PR to turn off the savings in this case if our goal is to speed up the full suite testing.

I'm tempted to undo my TEST_LONG_RUN changes so we can take advantage of the savings in the nightly suite.

@timmiesmith @SergeyKopienko What are your thoughts?

[danhoeflinger]

I'm tempted to undo my TEST_LONG_RUN changes so we can take advantage of the savings in the nightly suite.

I've reverted this change, because the point is to provide savings to the nightly suite. We can examine this further, perhaps we want some TEST_VERY_LONG_RUN option or something.

[SergeyKopienko]

LGTM