[FEA] Keyword argument support at kernel launch

[a-ellison]

Is your feature request related to a problem? Please describe.

Python supports passing function arguments either positionally or by keyword,
but Numba CUDA kernel launches do not appear to support keyword arguments.

Minimal repro:

from numba import cuda


def f(flag):
    return flag


@cuda.jit
def kernel_1(flag):
    return


@cuda.jit
def kernel_2(flag=True):
    return


print("plain Python call:", f(flag=True))

try:
    kernel_1[1, 1](flag=True)
except TypeError as exc:
    print(f"kernel_1: {exc}")

try:
    kernel_2[1, 1](flag=True)
except TypeError as exc:
    print(f"kernel_2: {exc}")

This fails with:

plain Python call: True
kernel_1: _LaunchConfiguration.__call__() got an unexpected keyword argument 'flag'
kernel_2: _LaunchConfiguration.__call__() got an unexpected keyword argument 'flag'

Describe the solution you'd like

Support keyword arguments as in regular Python. It's used often in Pythonic code.

Describe alternatives you've considered

N/A

Additional context

Observed with numba-cuda 0.30.0 and numba 0.65.0.