Fix capgen Fortran parser to recognize nested subroutines/functions

[climbfuji]

Fixes https://github.com/NCAR/ccpp-framework/issues/359 (https://github.com/NCAR/ccpp-framework/issues/359).

The capgen Fortran parser is updated to know whether it is inside a contains section of a CCPP scheme. If so, the contents is ignored until an end subroutine scheme_name_ccpp_phase (e.g. end subroutine mp_thompson_run) is found.

Note that this fix requires that end subroutine statements are labeled with the subroutine name, i.e. subroutine mp_thompson_run requires the end statement to be end subroutine mp_thompson_run and not just end_subroutine). This has been a soft requirement for ccpp_prebuild.py, which warns users that an unlabeled end subroutine statement was found, but in must cases was able to proceed.

A more complicated solution that does not require labeled end subroutine statements would require counting the number of [begin] subroutine and end subroutine statements inside contains regions.

User interface changes?: No (since labeled end subroutine statements have been a soft requirement for ccpp_prebuild.py from the beginning)

Fixes #359.

Testing: The existing test of scheme temp_calc_adjust was extended to include a contains section in temp_calc_adjust_run. Without the changes to scripts/fortran_tools/parse_fortran_file.py in this PR, the capgen test for this scheme failed. After making the changes suggested here, all tests pass (cd tests && ./run_test.sh).

[gold2718]

I'm not sure I believe your assertion that the end subroutine statement needs to label. I will submit a PR to your branch demonstrating this.