Add fsanitize testing

[aprokop]

Found some issues in #63. Plus we would want to satisfy an optional requirement of xSDK (cf #58).

[aprokop]

Actually, fsanitize is probably preferred to valgrind.

[masterleinad]

I just have problems running CMake using MPI then.

[dalg24]

You probably need a suppression file. Bruno had added one in DTK. Refer to ORNL-CEES/DataTransferKit@d2c8c4f

[masterleinad]

With -fsanitize=leak and surpressing errors for libopenpal.so(no CUDA), I currently see:

 SUMMARY: LeakSanitizer: 33065 byte(s) leaked in 407 allocation(s).
13:     #1 0x7fab0e05a547  (<unknown module>)
13:
13: Indirect leak of 80 byte(s) in 1 object(s) allocated from:
13:     #0 0x7fab12d80acb in __interceptor_malloc (/usr/lib/x86_64-linux-gnu/liblsan.so.0+0xeacb)
13:     #1 0x7fab0a158edc  (<unknown module>)
13:
13: Indirect leak of 56 byte(s) in 1 object(s) allocated from:
13:     #0 0x7fab12d80acb in __interceptor_malloc (/usr/lib/x86_64-linux-gnu/liblsan.so.0+0xeacb)
13:     #1 0x7f0dd9e75484  (<unknown module>)
13:     #2 0x3433  (<unknown module>)
13:
13: Indirect leak of 48 byte(s) in 2 object(s) allocated from:
13:     #0 0x7f0ddec05acb in __interceptor_malloc (/usr/lib/x86_64-linux-gnu/liblsan.so.0+0xeacb)
13:     #1 0x7f0431fa9c4c  (<unknown module>)
13:

[aprokop]

Apparently we are not doing it, it's been 3 years.

[dalg24]

Sure. We can always re-open as necessary. I have a suppression file laying around somewhere for stuff from Boost. I typically didn't use with MPI (mostly because I usually don't enable to save time when developing these days)