thrust::stable_sort_by_key has known issue with device linking. The code
crashes with cudaInvalidValueError. It works as expected without any
changes with or without separable compilation otherwise.

https://github.com/thrust/thrust/wiki/Debugging#known-issues
https://github.com/thrust/thrust/blob/master/doc/changelog.md#known-issues-2

The above documents mention a known issue with device linking and thrust.
Although the documents say it happens in debug mode(with -G flag), I noticed
similar crashes in release configuration too in ArrayFire.

Due to the above issue, I have separated out the relevant source files
(fft,blas,sparse and solver) which require device linking into separate
static library. Once separated into a separate static library, sort_by_key
and all the other unit tests that use it are running as expected without
any crashes.

Removed a special neighborhood iterator which isn't necessary

Added mouse manipulations

Change ninja to 1.10.0

pinverse_cpu test is excluded as lapacke dependency is not taken care of
yet

* Fix constant mem declaration in CUDA morph kernel

Global constant value of max filter length was not modified
after increasing filter support to 19 from 17 back originally.

* adds fallback for convolveNN functions

* adds cudnn option, runtime fallback

* Noexcept and const many Dependency module functions

* Refactor cuDNN code in CMake

* Fix fallback logic. refactor cuDNN util functions. Fix f16 wrap

Co-authored-by: Umar Arshad <umar@arrayfire.com>

* Add clang-tidy configuration file

* Cleanup some exception code

* Add additional upstream directories to .gitignore

* Remove unused parameters from wrap and transform implementations

* Fix warnings and removed unused calls

* Removed constexpr not supported by VS2015

* Fixed formatting

* Improve documentation of the alloc and free function
* Add tests for memory operations

* Created snippets for examples in the document

* The rdc and dlink flags are not required because they are added
  by CMake for separable compilation and static linking respectively
* Add guards around libs that are not included in the CUDA 9.0
  Toolkit
* Only link with OpenMP when linking with cuSOLVER dynamically
* Fix error message when CUDNN is not found

* Address casts from double to __half which are missing in 9.0
* Thrust return_temporary_buffer function can accept void* pointers
  in older versions of Thrust. Use raw_pointer_cast to pass the
  pointer to memFree
* cublasGemmEx doesn't exist in CUDA 9.0. Add ifdefs to guard
  against older builds
* __float2half is not a host function so it needs to be removed
  from mean
* Add template instantiation for memFree to accept void* pointers

CUSOLVER_CHECK error message printed "CUBLAS Error" instead of
CUSOLVER Error

Earlier to this change, I added bash based syntax which won't work
with /bin/sh or dash shells.

/usr/sh is available on most systems that use init.d scripts. So, it is
safe to assume it's availability on majority of linux distributions.

* Adds PR template 

**Short description of change**  
Adds a github PR template for the ArrayFire project. Developers will now face a short suggested checklist when creating a new PR on github.

**Motivation**  
Adding a PR template will make it easier to reference old issues when generating reports and link future issue in historical context.

**Future considerations**  
Wiki might need to be updated with additional development guidelines. The current guidelines could be more comprehensive.

* Updated pull request template

* Added additional detail.
* Use comments instead of text to communicate with the reader.
* Create a simple checklist

* Grammer + Future changes in the description section

Co-authored-by: Umar Arshad <umar@arrayfire.com>

* AF_CONSTEXPR expands to nothing if constexpr support is not available.

* Replace CONSTEXPR_DH with AF_CONSTEXPR and __DH__ in
  `src/backend/common/half.hpp`

* Removed AF_CONSTEXPR where it is invalid in half.hpp

* Adds the Zc:__cplusplus flag to cuda builds for MSVC if the flag is available.
the cuda_fp16 header does not define the default constructor for __half as
"= default" and that prevents the __half struct to be used in a constexpr
expression

* For older versions of MSVC we define the __cplusplus macro before and
after the inclusion of cuda_fp16.h header.

* Define the AF_CONSTEXPR macro for NVRTC compilation

Adds several classes of issues with proposed additional information that would be helpful when debugging.

Co-authored-by: pradeep <pradeep@arrayfire.com>
Co-authored-by: Umar Arshad <umar@arrayfire.com>

cusparseSpMv/cusparseSpMM functions use sparse and dense matrix/vector
descriptor objects as arguments. This API is introduced in CUDA 10.1 and
old API has been deprecated. It is also removed in CUDA 11.

Also, updates CUB version from 1.8.0 to 1.9.10