Enhance querying kernels preferred wgsize

[omarahmed1111]

Work-group sizes currently rely on device maximum rather than the max from a kernel query. This could result in an error raised as the device maximum could be more than what the kernel is actually allowed to use.

This PR uses an approach to make choosing the wgsize more safer for the kernels. The approach used composed of 2 sides:

• if the reduction kernel was given a name by the user parallel_for<class Name> then we use this name to query the best wgsize for this kernel.
• If the reduction kernel is not name defined by the user, we use an approximate safe approach where we query all the reduction kernels in the sycl application for their best wgsize, and we pick the minimum wgsize and use it for the kernel.

The second approximate approach part could be more accurate by using this PR that would give each reduction kernel a unique name that would make querying them possible at runtime.