Check all basic operators results in all possible cases

[GillesDuvert]

added test_indepth_basic_functions.pro and modified code to insure ~1200 tests on OPERATORS are similar to IDL.

[codecov]

Codecov Report

Attention: Patch coverage is 78.04054% with 65 lines in your changes are missing coverage. Please review.

Project coverage is 43.73%. Comparing base (2980154) to head (eb9e313).

Files	Patch %	Lines
src/basic_op_new.cpp	59.01%	25 Missing ⚠️
src/basic_op.cpp	70.21%	14 Missing ⚠️
src/snippets/basic_op_GtMarkNewCplx.incpp	60.00%	4 Missing ⚠️
src/snippets/basic_op_LtMarkNewCplx.incpp	60.00%	4 Missing ⚠️
src/snippets/basic_op_AndOpCplx.incpp	57.14%	3 Missing ⚠️
src/snippets/basic_op_AndOpInv.incpp	57.14%	3 Missing ⚠️
src/snippets/basic_op_LtMarkSNewCplx.incpp	83.33%	2 Missing ⚠️
src/snippets/basic_op_AndOpInvS.incpp	90.90%	1 Missing ⚠️
src/snippets/basic_op_DivInvSCplx.incpp	91.66%	1 Missing ⚠️
src/snippets/basic_op_DivInvSNew.incpp	92.30%	1 Missing ⚠️
... and 7 more

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1784      +/-   ##
==========================================
+ Coverage   42.70%   43.73%   +1.02%     
==========================================
  Files         361      376      +15     
  Lines       97331    97398      +67     
  Branches    19928    19919       -9     
==========================================
+ Hits        41570    42596    +1026     
+ Misses      55761    54802     -959     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[GillesDuvert]

The test_indepth_basic_functions.pro brutally explores 95 different operators defined in files basic_op_xxx.pro, for a combination of the following cases:

1. The 10 different numeric operand types (BYTE to DCOMPLEX)
2. The size of each operand: scalar, array of 1 element, array of > 1 element
3. The respective sizes of the two operands for operators using two operands (like in c= a + b ), the code is not the same when a is smaller of greater than b.
4. The persistence of the operand, like in c = a + temporary(b)

Because each of these cases need a special code, for speed and memory optimization (this was written by Marc in the very first version of GDL, at least for 90% of the operators), the number of combinations explored is around 11000 cases.

The result of the procedure is a ~24000 lines text file that can be compared with an equivalent IDL output.
The present PR introduces a number of code patches or additions to insure even obscure operator+operand combination give the same result as IDL.
The GDL and IDL produced files are identical, with the exception of the signedness of a few NaN or Inf results in case or zero division of complex values.

As each operator is also optimized to use parallel processing on operands as soon as the number of elements is greater than !CPU.TPOOL_MIN_ELTS, the test_indepth_basic_functions.pro procedure can also be used to check that the parallelized results are identical to the single cpu ones (*). This has been checked too. Note that testing the equivalence of results between single and many cpu as performed in the test_indepth_basic_functions.pro procedure with IDL will... crash IDL on some complex values, so this part of procedure is deactivated with IDL.

(*) TPOOL (Thread Pool) using functions cannot in general be expected to provide exactly the same result in parallell or not parallell mode: TOTAL() is an example. But operators do.