Adding bignum in to the benchmarks.

[Machiry]

Added a new benchmark (bignum).
The workflow run corresponding to this: https://github.com/correctcomputation/actions/runs/3284592753?check_suite_focus=true

[mattmccutchen-cci]

Two issues with this benchmark that are not in the PR itself:

1. I found some diffs between the tar file on gamera and the original tiny-bignum-c. Did you make these changes yourself, or where did they come from?

2. In the workflow run, I noticed that the post-conversion build stopped after the failure in golden. In general, we want the post-conversion build to attempt all translation units and report all errors so that we get meaningful counts in summarize-benchmark-errors.py. I looked at the makefile, and it is weird: it has a single all rule that runs all the compilation commands in sequence, and it compiles bn.c multiple times, so warnings would be reported multiple times (if we ever track warnings in addition to errors). I'd like to replace the build rules with something saner; this will fix the error reporting and also allow the build to run in parallel for a slight speedup. Here's a draft of the new build rules I'm thinking of. Does this seem reasonable to you? If so, what would be the best way to incorporate the change into the workflow? Put a patch file in /home/github/checkedc-benchmarks on gamera and have the workflow apply it?

[mattmccutchen-cci]

Suggested change

	# bignum
	# tiny-bignum-c

Can we please refer to this benchmark consistently as tiny-bignum-c (also in the PR title)? If you say "bignum", it's not obvious that you're talking about the same thing because there are many pieces of software with similar names on the web.