Cartesian indexing for SubArrays with non-integer AbstractRanges (#52094)

With #50457 now merged, an `AbstractRange` of `CartesianIndex`es should
use `CartesianIndexing` in constructing a `SubArray`. I've limited the
method to integer ranges, as this is the case where we know for sure
that the indexing may be linear.

Fixes
```julia
julia> view(1:2, StepRangeLen(CartesianIndex(1), CartesianIndex(1), 0))
0-element view(::UnitRange{Int64}, StepRangeLen(CartesianIndex(1,), CartesianIndex(1,), 0)) with eltype Int64
```

---------

Co-authored-by: N5N3 <[email protected]>

Author: jishnub

base/subarray.jl

@@ -52,8 +52,10 @@ viewindexing(I::Tuple{Slice, Slice, Vararg{Any}}) = (@inline; viewindexing(tail(
 # A UnitRange can follow Slices, but only if all other indices are scalar
 viewindexing(I::Tuple{Slice, AbstractUnitRange, Vararg{ScalarIndex}}) = IndexLinear()
 viewindexing(I::Tuple{Slice, Slice, Vararg{ScalarIndex}}) = IndexLinear() # disambiguate
-# In general, ranges are only fast if all other indices are scalar
-viewindexing(I::Tuple{AbstractRange, Vararg{ScalarIndex}}) = IndexLinear()
+# In general, scalar ranges are only fast if all other indices are scalar
+# Other ranges, such as those of `CartesianIndex`es, are not fast even if these
+# are followed by `ScalarIndex`es
+viewindexing(I::Tuple{AbstractRange{<:ScalarIndex}, Vararg{ScalarIndex}}) = IndexLinear()
 # All other index combinations are slow
 viewindexing(I::Tuple{Vararg{Any}}) = IndexCartesian()
 # Of course, all other array types are slow

test/subarray.jl

@@ -809,3 +809,11 @@ end
     b = view(a, Base.IdentityUnitRange(4:7))
     @test first(b) == a[Base.first_index(b)]
 end
+
+@testset "StepRangeLen of CartesianIndex-es" begin
+    v = view(1:2, StepRangeLen(CartesianIndex(1,1), CartesianIndex(1,1), 0))
+    @test isempty(v)
+    r = StepRangeLen(CartesianIndex(1), CartesianIndex(1), 1)
+    v = view(1:2, r)
+    @test v == view(1:2, collect(r))
+end