refactor inlist reading

[VincentVanlaer]

This commit introduces utils_namelist, which abstracts some of the common aspects of reading inlists in MESA (error message, nested inlists, ...). Previously, various different namelist reading routines were copied and modified around the code base. These have accrued various differences over time, which has been fixed now.

The following behaviour has been changed:

• pgbinary and pgstar no longer make MESA error when they are missing from inlists
• it no longer matters where in the chain of inlist a namelist section is missing. It used to be that for certain section, only the first inline in a chain was allowed to have a missing section.
• checks and copying of options only happens once all inlists have been read
• failures when reading inlists will no longer dump a stack trace on the user

[VincentVanlaer]

The way I have implemented this is different from the usual way MESA is structured. Rather than having the utils_namelist module be included with utils_lib, I have exposed it separately. The main reason for doing is that it allowed me to iterate quicker on the interface design, as changes to the utils_lib interface require most of MESA to be recompiled.

[warrickball]

I just tried to run this through mesa_test on our cluster and everything failed with messages like:

 [ERROR]: too many levels of nested star_job inlist files
 ierr from read_star_job inlist
rn1 failed
******************** pisn failed: non-zero exit code ********************

where you can replace pisn with any other test case.

I doubt I'll have time to investigate before you do but if you try to run any test case in your dev version you should be able to rule out this error. If you have already managed to run any test OK, let me know.

FWIW that's on our cluster running RHEL 8.10 using MESA SDK 26.3.2.