* CI speedup and simplification

* Rebuild cached master when branch build fails

* Fixed sed command

* Fixed cahced nGraph path

* Exit function if cmake or make fails

* Fixed build path

* Fail building in case cached ngraph is not built

* Ensure cached nGraph is on master before pulling

* add ngraph dequantize op

* use a floating point offset

* code format

* reminder to fix serializer

* add serializer support

* add dequantize test cases

* cleanup and code format

* fix build warning for implicit conversion

* remove get_tensor method

* remove get_element_count as it is redundant to get_size

* fix a few issues with examples and python wrapper

* rename get_size to get_element_count and add get_size_in_bytes method

* add quantize op into intelgpu

* IntelGPU backend: ArgMax, ArgMin, LRN operations

* PR1697. Comments addressed

* PR1697. Empty lines added

* add DropOut op and stub to Documentation ToC

* cleanup stub for dropout and add template for op additions

* add checklist as helper for op_stub

* adding nGraph Quantize op

* unit test failing for floating point exception

* unit test working in float

* unit test working in uint8

* improved type checking and polished unit test - passing

* quantized axes working

* inclusive project method

* add round mode

* TODO cleanup

* code format

* adding serializer support - fails build

* add serializer support

* make CPU quantize op work; new tests for int8, clamp)

* fix build failure

* fix GPU build issue

* fix GPU unit test manifest

* use quantized offset

* add is_quantized field to element::Type

* add reduce function to coordinate.hpp

* Move CallFrame functionality into GPUBackend

* IntelGPU backend: Switch Select operation to use implementation from clDNN

* PR1688: Style fix.

Fix DEX_ONLY compile error in cpu_external_function.cpp

Move definition of StaticInitializers out of `#ifndef NGRAPH_DEX_ONLY`

* IntelGPU backend: Types workaround removal

* PR1681. Fix style

* Don't reuse memory for non-destructive ops.

* Update comment.

* CPU Builder: Replace uint8_t with double in partial element type set

* Keep support for uint8_t kernels

* update nbench output to be more consistent for single vs. directory benchmark.

* wip

* add types used in model

* fix test

* style

* catch all exceptions when running any option

* fix CPU dependency

* zero dim elem fix

* switch to find

* fix runtime errors

* IntelGPU backend. Use common nGraph optimization passes

* PR1667. Move variable definition.

* Add CPU horizontal fusion pass for inception.

* Name change.

* Move horizontal fusion to cpu_fusion.

* Change horizontal fusion pass for inception to a general horizontal fusion pass.
Add a unit test conv_horizontal_fusion to cpu_fusion.

* Rename files.

* Correct cpu_fusion.hpp.

* Add NGRAPH_DEBUG.

* Set native layout when input format of slice is nChw16c or nChw8c and lower bound of
channels is not a multiple of 16 or 8.

* added stream writers to dump meta information related to CONSTANT, INTERMEDIATE, PARAMETER's

* WIP added method to dump Input and output reference address in to debug manifest from cpu_call_frame

* added suppport to dump memory refernce in the debug manifest

* - added macro to dump debug manifest
- added support to release_function once dumping the memory references

* added a method to write a generated stream to a file

* dont release function, if NGRAPH_DEX_DEBUG flag set

* Addressed PR comments
-   dump debug manifest using stringstream instead of CodeWriter
-   Moved dumping debug manifest logic to cpu_external_function

* revert changes to call_frame

* - Addressed PR comments

* :Remove call_once from dumping debug_manifest

* address PR comments

* address PR comments

* extend concat elimination to fold slice + reshape

* relax branch tip to be anything not just goe

* add support for transpose when concat of slice + reshape occurred on internal axis.

* simplify reshape order permutation

* multi-axis slice + concat do not cancel and are now disabled. generalize detection of axis reordering when intermediate reshape is present
to include logical reshape and reshape that results in axis reordering for the cases: parent_shape.size gt, eq, and lt concat_shape.size.

* check that slices are in order

* add one user check on reshape

* add more checks

* fix warnings

* Reshape axis order did not include enough dimensions when the transposed and reshaped result was of lower rank.

* inject details into visualize_tree pass

* address bob's feedback

* revert back to map

* - fixed performance_counter for CPU DEX

* made changes to start clock @ the first functor and collected perf_counter @ the exceution of last functor for given kernel execution

* - made changes to have max_pool kernel to have exactly one functor
- assert if functor != num_of_ops

* fixed profiler_counter for TBB execution

* avoid unnecessary calculation to enables<map> value

* Addressed PR comments
-   updated "enables" -> list<functors>

* Common pass registration for codegen and Dex

* Make return indices optional for cpu workspace insertion

* [ONNX] LRN op

* Review fix pt. 1