[SYCL][UR] Move `max_global_work_groups` query to UR (#21840)
This PR makes the following changes:
1. Moves implementation of `max_global_work_groups` to UR. This query
has been implemented in SYCL RT because there's no backend support for
`max_global_work_groups` query. However, it was recently decided that
OpenCL will add a corresponding query. See CMPLRLLVM-73572 for more
info.
2. Changes `max_global_work_groups` from `INT_MAX` to `SIZE_MAX` for all
backends. For CUDA, HIP, OFFLOAD, and L0 adapter, we calculate the value
of `max_global_work_groups` by taking minimum of `SIZE_MAX` and
multiplication of per-dimension max group size.
3. Changed `max_work_groups<3>` so that `max_global_work_groups` no
longer limits per-dimension max work group size.
---------
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Co-authored-by: iclsrc <iclsrc@intel.com>
