DimShuffle's Numba implementation cannot handle cases when output is empty/scalar.

[kc611]

Description of your problem or feature request

DimShuffle's Numba implementation fails when output shape is () (i.e. no output when all dimensions are dropped). This happens because of the n > 0 restriction placed on create_tuple_creator:

aesara/aesara/link/numba/dispatch/basic.py

Lines 218 to 223 in e6c2fbc

	def create_tuple_creator(f, n):
	"""Construct a compile-time ``tuple``-comprehension-like loop.
	See https://github.com/numba/numba/issues/2771#issuecomment-414358902
	"""
	assert n > 0

Which is used in the implementation of DimShuffle to create output 's shape tuple:

aesara/aesara/link/numba/dispatch/elemwise.py

Line 327 in e6c2fbc

create_zeros_tuple = numba_basic.create_tuple_creator(lambda _: 0, ndim_new_shape)

Please provide a minimal, self-contained, and reproducible example.

import numpy as np
import aesara
from aesara.graph.optdb import OptimizationQuery
from aesara.compile.mode import Mode
from aesara.link.numba.linker import NumbaLinker
from aesara.tensor.elemwise import DimShuffle
from aesara.tensor.type import TensorType

opts = OptimizationQuery(include=[None], exclude=["cxx_only", "BlasOpt"])
numba_mode = Mode(NumbaLinker(), opts)
py_mode = Mode("py", opts)

orig_shape = (1, 1, 1)
dims = ()
out_shape = ()

broadcastables = [(entry == 1) for entry in orig_shape]
x = TensorType(aesara.config.floatX, broadcastables)("x")
out = DimShuffle(broadcastables, dims)(x)

py_res = aesara.function(inputs=[x], outputs=[out], mode=py_mode)(np.ones(orig_shape, dtype=aesara.config.floatX)) # Passes
numba_res = aesara.function(inputs=[x], outputs=[out], mode=numba_mode)(np.ones(orig_shape, dtype=aesara.config.floatX)) # Fails
pass