PROJ/test/cli/testntv2, branch 6.3

Database: tune accuracy of Canadian NTv1 file w.r.t NTv2

2019-12-25T15:23:31+00:00

As on import of EPSG, we remove the supersession of Canadian NTv1 file w.r.t NTv2
(because the default behaviour of PROJ is to ignore superseded operations). However
the NTv1 operation is advertized with an accuracy of 1m, whereas NTv2 is advertized
with 1.5m. Consequently on areas where both files are valid, and if both files are
available, NTv1 would be selected. So as a workaround, worsen the NTv1 accuracy to
2m so that NTv2 is used in priority.

proj_trans: add retry logic to select other transformation if the best one fails.

2019-12-24T11:59:15+00:00

Relates to https://github.com/OSGeo/PROJ/issues/1808

Horizontal grid shift: fix issue on iterative inverse computation when switching between (sub)grids (fixes #1663)

2019-12-14T14:40:17+00:00

Given in.txt with
53.999759140 5.144478208 252.6995

Before the fix,
cct  -t 0 -d 4 +proj=pipeline +step +proj=axisswap +order=2,1,3,4 +step +proj=hgridshift +inv +grids=rdtrans2018.gsb +step +proj=vgridshift +grids=naptrans2018.gtx +step +proj=sterea +lat_0=52.156160556 +lon_0=5.387638889 +k=0.9999079 +x_0=155000 +y_0=463000 +ellps=bessel in.txt

returned:
  139079.8814    668306.0302      212.1724        0.0000

It now returns:
  139079.8850    668306.0458      212.1724        0.0000

which meets with the 1mm accuracy the expected result of test point
```
30010049    53.999759140    5.144478208     252.6995        139079.8850     668306.0460     212.1723
```

Adjust test outputs

2019-04-05T04:47:15+00:00

 - Tests should use unified diff with expected results
 - Show CMake build compile commands (VERBOSE=1)
 - Travis CI only has two cores, so "-j2" is sufficent

Fix data path used by cli tests (#1288)

2019-02-24T13:02:45+00:00

All other tests use PROJ_LIB, and allow it to be overridden from the
command-line, so do the same here.

Rename test/old/ to test/cli/ to better reflect their nature

2018-12-01T18:15:48+00:00