parallel support removed from default compiler options

The default compiler settings (stored in "setup_gcc.sh" and "setup_clang.sh") generate a binary which no longer runs in parallel (OpenMP). I made this change in response to issue #1. Although the program appears to be working again now with all arguments and all compiler settings, I felt that the additional speed from OpenMP did not justify increasing the doubt in the program's reliability. For most use case scenarios, this program is already fast enough.

Avoiding the use of OpenMP may slow down the program somewhat. If this is a problem, compile the program using the "setup_gcc_parallel.sh" or "setup_clang_parallel.sh" compiler options and pay attention to the output that the program generates. (Fortunately, if you do run into that bug, it will be obvious there is something wrong.)

now works again with OpenMP

Found and fixed a bug that was preventing ndautocrr from running correctly when OpenMP was in use (as it is by default). (My apologies. The previous version of this program was only tested in serial.)

improved the default behavior and fixed "-t"

There are 3 changes in this release

1. changed the default truncation length L to something more sensible
2. corrected a bug in the "-t" argument
3. improved the way correlation lengths are estimated and printed to the standard error. (This new way should be more robust.)

As a result, now you can run ndautocrr without any arguments and it should behave sensibly and run quickly.

initial release

This initial release has the basic features working to support the moltemplate examples which depend on it.

It's possible that when multiple different data sets are supplied, the program may be computing the average in a suboptimal way. (The average is computed independently for each data set, instead of over all data sets. This would probably to cause the program to underestimate correlations, including correlation times and lengths.) But this is not an issue if you use the -avezero command line argument. (Currently the only moltemplate example which depends on this code uses this argument, so this potential criticism is not relevant.)