Use dynamic allocation with bufr for all compilers

[mathomp4]

This PR turns on DYNAMIC_ALLOCATION for all compilers in bufr.

[gmao-msienkie]

This would be useful for some future work that Josh will be doing with high resolution radiosondes, since that will require that he compile the GSI with dynamic allocation.
There may be some other changes (i.e. bugfixes) in the BUFR library that would be useful to pull in. I'll have to look at what I put in my copy of the BUFR library when I was working with the high resolution sondes. I would not recommend going all the way to the latest version, or at least not before making sure that any r8i8 code (Chris Redder's radcor, perhaps) has been modified to r4i4.

[mathomp4]

As far as I know, it should be zero-diff. If nothing else, it has to be used eventually!

[gmao-msienkie]

I think that the idea here is to make sure that something doesn't break. I know you said there was a change to have it print a warning message rather than crash if the code tried to closbf before openbf but that might mean that there are a zillion warning messages printed. One of the changes for hdraob in my branch feature/msienkie/hdraob_mods includes an 'openbf' call in glbsoi which apprarently takes care of this issue, and maybe we could use that instead of having to change a lot of code with 'closbf' statements in wrong locations.
I have no idea whether the preprocessing QC under NCEP_Paqc has any trouble with the dynamic allocation but I can test those programs sometime soon - I am working on a patch for one of the programs in that section.