Fix TypeError when reporting ParseSyntaxError in parse_fortran_var_decl

[jimmielin]

[ 50 character, one line summary ]
Fix TypeError from ParseSyntaxError in parse_fortran_var_decl

[ Description of the changes in this commit. It should be enough
information for someone not following this development to understand.
Lines should be wrapped at about 72 characters. ]
In parse_fortran_var_decl(), when FortranVar() raises a
ParseSyntaxError, the exception object was appended directly
to the errors list. parse_module expects string elements and calls
'\n'.join(errors), which crashes with a TypeError. Convert
the exception to a string before appending.

From the context above errmsg is already being appended to
errors array so I think this is a safe change to make that is
consistent with the intended type of errors[].

User interface changes?: No

Fixes: closes #725

Testing:
test removed:
unit tests:
system tests:
manual testing: see minimal reproducible example in issue #725

[nusbaume]

The source code change looks good to me, but should we add some sort of doctest or unit test to ensure that the error you ran into will continue to be caught correctly in the future?

[jimmielin]

Thanks @nusbaume for the suggestion, I added a doctest that should cover this change
(the other doctest on the line above it doesn't go to the ParseSyntaxError path so it was not covering this bug)

If the change is not applied, this new test would fail:

702     >>> parse_fortran_var_decl("type(banana_t) :: bananas(0:N_FRUITS)", ParseSource('foo.F90', 'module', ParseContext()), _DUMMY_RUN_ENV)[1][0]
Expected:
    "bananas: '0:N_FRUITS' is an invalid dimension name; integer dimension indices not supported, in <standard input>"
Got:
    ParseSyntaxError("bananas: '0:N_FRUITS' is an invalid dimension name; integer dimension indices not supported, in <standard input>")

[nusbaume]

Thanks for the new test @jimmielin! Everything looks good to me now (albeit as a non-framework expert).