[Hexagon] Add tests for 2 operators that utilize Qualcomm HVX intrinsics.

[nverke]

Background This PR adds 2 tests for unused HVX implementations of mmul and conv2d. These were created to experiment with the performance of HVX intrinsics in tvm and each test outputs the runtime of each on various different input shapes. These will also be a good reference point for others looking to start using HVX intrinsics in their schedules.

Testing There are no code changes here outside of the tests themselves.

cc @mehrdadh