[GPU] Add a fused kernel groupnorm implementation

[mdvoretc-intel]

Details:

• Add a new OCL implementation for fsv16 group normalization
• The new implementation is used if each group contains fewer than fsv=16 features
• A single fused kernel handles all stages of the reduction, avoiding excessive loading of shared values and reusing cache in cases of small inputs

Tickets:

• CVS-177817

[e-ddykim]

Please add tests.

[mdvoretc-intel]

group_normalization.basic_b_fs_yx_fsv16 is currently selecting the new implementation. This is not intended behavior and interferes with keeping a separate specific test for the existing fsv16 implementation. While there is a selection between the two in other tests, this still needs to be resolved as part of test addition.