Improve test data structure

[bpurinton]

Currently, fixture test data is stored in the repository in tests/test_data/. That's probably not best practice, but it sure is convenient...

Some inspiration can be found in the xdem and geoutils packages.

---

The ASP uses this tarball for running tests (568 MB size):

wget https://github.com/NeoGeographyToolkit/StereoPipelineTest/releases/download/0.0.1/StereoPipelineTest.tar

As specified here, it contains all tests scripts, data, and expected results.

The data/ directory in that tarball looks like:

8.9M - 13APR25_WV02_SEVZ_1030010021A8A100_10030010021A64500_DEM3-3m_10pct.tif
 55M - B17_016219_1978_XN_17N282W.8bit.cub
 37K - B17_016219_1978_XN_17N282W.8bit.json
 55M - B18_016575_1978_XN_17N282W.8bit.cub
 41K - B18_016575_1978_XN_17N282W.8bit.json
3.1M - Copernicus_DSM.tif
120M - M181058717LE.ce.cub
171K - M181058717LE.ce.json
120M - M181073012LE.ce.cub
176K - M181073012LE.ce.json
6.3M - Severnaya-Bedrock-UTM47-Ellipsoidal-Height.txt
5.5M - basic_panchromatic.tif
 20M - ref-mars-PC.tif
147K - ref-mars_yxz.csv

Can we use that data in the tests here as well? Or perhaps better to keep separate, some useful files maybe? (But probably not the .cub or .json files).