[Coverity] Review kernel program cache locks

You can see Coverity issues at https://scan.coverity.com/projects/intel-llvm?tab=overview

This report is motivated by `512040` - from what I understand, Coverity's concern is that here we acquire multiple locks and it may lead to a deadlock if there is another place where the same set (or a subset bigger than 1 lock) is acquired as well, but in a different order.

I don't know if we have a precedent for this, but the comment around the function says that it should only be used in unit tests.
That makes me think that it may not be ready for a production usage in a multi-threaded environment.

https://github.com/intel/llvm/blob/98cd46434d99fe9bbca7bf7ca104e1164939b97f/sycl/source/detail/kernel_program_cache.hpp#L716-L722

However, it is used several times in the SYCL RT:

https://github.com/intel/llvm/blob/98cd46434d99fe9bbca7bf7ca104e1164939b97f/sycl/source/detail/kernel_program_cache.hpp#L811

https://github.com/intel/llvm/blob/98cd46434d99fe9bbca7bf7ca104e1164939b97f/sycl/source/detail/program_manager/program_manager.cpp#L1813

https://github.com/intel/llvm/blob/98cd46434d99fe9bbca7bf7ca104e1164939b97f/sycl/source/detail/program_manager/program_manager.cpp#L3008

I think that we should clarify the comment and double-check if a deadlock Coverity is talking about is indeed possible or not