Fix array conversion

[urschrei]

My suspicion is that creating the raw pointer in the function call was causing it to be dropped prematurely, resulting in garbage data. See here for an analogous issue when getting a CString pointer. We avoid this by explicitly creating the pointer before the call. In addition, the doctest has been updated with a less complex example that doesn't use a proj string.

[urschrei]

Bors try

[bors[bot]]

try

Build succeeded:

• ci result

[michaelkirk]

bors try

Interesting! Let's gather some stats here.

[urschrei]

If your run passes, I'm gonna contrive a doctest with different CRS and coordinates. I have my doubts about the validity of the old Stereo70 conversion values, but there's less value in a doctest that just reproduces an existing known-good test.

[bors[bot]]

try

Build succeeded:

• ci result

[urschrei]

bors try

[bors[bot]]

try

Build succeeded:

• ci result

[urschrei]

These are CRS and coordinate values from https://geodesy.noaa.gov/NCAT/.

WDYT?

[urschrei]

bors r=michaelkirk

[bors[bot]]

Build succeeded:

• ci result