Tracking NumPy usage in various modules.

[kc611]

Feature request

As titled, this PR tracks usage of NumPy across Numba. The end goal is to hopefully refactor NumPy usage in a way such that it is restricted to either numba.np and/or numba.cuda module in order to simplify dependencies.

Following are places other than numba.np where NumPy is being used:

• numba.core
  • boxing.py (Will require type separation between NumPy and Python scalar types.)
  • errors.py
  • inline_closurecall.py
  • ir_utils.py
  • utils.py
  • datamodel.models.py
  • types module (Will require type separation between NumPy and Python scalar types.)
  • typing module (Will require type separation between NumPy and Python scalar types.)
• numba.cpython
  • builtins.py
  • charseq.py
  • hashing.py
  • mathimpl.py
  • numbers.py
  • randomimpl.py
  • unicode_support.py
  • unicode.py
• numba.misc
  • dummyarray.py
  • gdb_print_extension.py
  • mergesort.py
  • numba_gdbinfo.py
  • numba_sysinfo.py
  • quicksort.py
  • special.py

For the .parfors, .stencils and .pyccmodules, they can be considered their own separate specialized modules depending on the numba.np module.

Moreover there are following QoL tasks:

• A test to make sure we don't accidentally add more NumPy usage in modules other than numba.np