Slice indexing of python objects?

I'm trying to use [GridSpec](https://matplotlib.org/api/gridspec_api.html#matplotlib.gridspec.GridSpec) in matplotlib with PyCall, and ran into this error:

```julia
julia> using PyCall, PyPlot

# Make a GridSpec object
julia> gs_main = PyPlot.matplotlib[:gridspec][:GridSpec](1, 2)
PyObject

# In python, you can use slice indexing on a GridSpec object to make subplots
# However, the naive way of going about this doesn't work:
julia> gs_main[1:2]
ERROR: MethodError: no method matching getindex(::PyObject, ::UnitRange{Int64})
Closest candidates are:
getindex(::PyObject, ::Symbol) at /home/glynch/.julia/packages/PyCall/rUul9/src/PyCall.jl:294
getindex(::PyObject, ::AbstractString) at /home/glynch/.julia/packages/PyCall/rUul9/src/PyCall.jl:283
getindex(::PyObject, ::Integer) at /home/glynch/.julia/packages/PyCall/rUul9/src/PyCall.jl:755
...
Stacktrace:
[1] top-level scope at none:0

# Manually indexing with slice objects does work, however
julia> gs_main[:__getitem__](pybuiltin(:slice)(0, 2))
PyObject

```

Is there a more straight-forward way of accomplishing this?

Reproduce the GridSpec example from the issue, then inspect the getindex methods in PyCall.jl around lines 283, 294, and 755. Compare the failing Julia range expression with the working __getitem__ call using pybuiltin(:slice); done means an idiomatic range-based index works for this Python object without breaking the existing symbol, string, or integer cases.

Vom Indexierungsmodell aus dem Issue-Text verfasst.