2 bugs in ERODE(), one to be fixed ...

[alaingdl]

I made some clean-up in "test_dilate.pro" and "test_erode.pro", should appear in current PR

As a result, I found 2 problems in ERODE() :

• the second parameter (structure) may changed depending it relation with the dims of first parameter (image). I solved that in sr/pro/erode.pro
• when the second parameter is a vector, we do have a shift in output, which is detected for odd and even inputs
  (see TEST_ERODE_AXES in "test_erode.pro")

I also realize that this test was deactivated then the problem was forgotten.

See also #461
Important point : we still do have internal versions of these 2 codes but it is not useful for my case :(

 GDL> help,/lib, name='*erode*'
Library procedures (0):
Library functions (1):
res=ERODE_INTERNALGDL([5 Args],HELP,GRAY,PRESERVE_TYPE,UINT,ULONG,VALUES)
GDL> help,/lib, name='*dilat*'
Library procedures (0):
Library functions (1):
res=DILATE_INTERNALGDL([5 Args],HELP,GRAY,PRESERVE_TYPE,UINT,ULONG,VALUES,CONSTRAINED,BACKGROUND)

[GillesDuvert]

@alaingdl I let you squash/merge your PR and tackle this!