[GPU] Support for 3D weights format for onednn.

[jade-cho]

Details:

• The failure occurred because a 4D tensor was converted into a 3D tensor for oneDNN primitives, and the cldnn::format::ioyx format was not properly handled during this conversion. As a result, the primitive failed to be created.
• Added a function to correctly handle this format (ioyx) along with other formats that could potentially cause similar issues during the conversion process.

The code and line that caused this issue

• openvino/src/plugins/intel_gpu/src/graph/impls/onednn/utils.cpp

  Lines 358 to 364 in 0e3f77a

  	dims.push_back(std::max(l.spatial(0), l.spatial(1)));
  	if (l.get_format() == format::bfyx)
  	target_fmt = dnnl::memory::format_tag::abc;
  	else if (l.get_format() == format::byxf)
  	target_fmt = dnnl::memory::format_tag::acb;
  	else
  	OPENVINO_THROW("[GPU] Unexpected layout format " + l.to_short_string());

Checklist

• Is it a proper fix? (not a workaround)
• Did you include test case for this fix, if necessary?
  • The unit tests will be added with another PR. Passed e2e test.
• Did you review existing test that can be extended to cover this scenario? Which test did you review?

Tickets:

• 177581

[hyunback]

LGTM, but do we need to add test code for problem case?

[wilson-seok]

LGTM

[isanghao]

LGTM