[GPU] Broadcast-Quantize fusion support (#33602)
### Details:
- *Fixed the `use_opt_code` execution path in `broadcast_gpu_ref.cl` by
adding the missing iteration over the `y` dimension in the post-ops
kernel.*
- *Added support for the quantize post-op within the broadcast GPU
reference kernel.*
- *Extended `BroadcastFusingTest` with new test instances that validate
the above changes.*
- *Added a new failing test case to `EltwiseFusingTest` (`eltwise →
i8_quantize → abs`). This issue was discovered during broadcast kernel
debugging but is not directly related to the broadcast implementation.*
### Tickets:
- *CVS-175802*
