[IE CLDNN] Implement EmbeddingBag operations

[echuraev]

Implemented three operations: EmbeddingBagPackedSum,
EmbeddingBagOffsetsSum and EmbeddingSegmentsSum. These operations do
the same work but have a different format of inputs.

Details:

• EmbeddingBagPackedSum - This is EmbeddingBagOffsetsSum with "bags"
  of equal size. If 3rd input is not provided, this operation is
  equivalent to Gather followed by ReduceSum(axis=0). However,
  EmbeddingBagPackedSum is much more time and memory efficient than
  using a chain of these operations.
• EmbeddingBagOffsetsSum - For each index in indices this operator
  gets values from data embedding table and sums all values belonging
  to each bag. Values in offsets define starting index in indices
  tensor of each "bag", e.g. offsets with value [0,3,4,4,6] define 5
  "bags" containing [3,1,0,2,n-6] elements.
• EmbeddingSegmentsSum - For each index in indices this operator gets
  values from data embedding table and sums all values belonging to
  each segment. Values in segment_ids define which segment index in
  indices tensor belong to, e.g. segments_ids with value
  [0,0,0,1,1,3,5,5] define 4 non empty segments other segments are
  empty, the number of segments is defined by num_segments input.

JIRA: 29994

[echuraev]

Please don't see on the first commit [WILL BE REMOVED] Changes in ngraph. I'll remove this commit when PR #576 will be merged to master. I didn't find how can I set on the github that PR is dependent on another PR.