finalization segfaults when flint built against older NTL than M2

[d-torrance]

There's currently an interesting problem in Debian. NTL 11.5.0 was recently uploaded to Debian unstable, and the latest version of the Macaulay2 package was built against it. However, the flint (and factory) versions that are currently in Debian testing were built against NTL 11.4.3.

Mixing these NTL versions causes a segfault when finalizing after doing things with Galois fields, e.g.,:

i1 : GF 8

o1 = GF 8

o1 : GaloisField

i2 : exit
-- SIGSEGV
-* stack trace, pid: 3744399
 0# stack_trace(std::ostream&, bool) at ./M2/Macaulay2/d/main.cpp:124
 1# segv_handler at ./M2/Macaulay2/d/main.cpp:241
 2# 0x00007FB6B00DAEF0 in /lib/x86_64-linux-gnu/libc.so.6
 3# NTL::UniqueArray<NTL::UniquePtr<NTL::FFTPrimeInfo, NTL::DefaultDeleterPolicy> >::~UniqueArray() at ../include/NTL/SmartPtr.h:1593
 4# __cxa_finalize in /lib/x86_64-linux-gnu/libc.so.6
 5# 0x00007FB6AF9515B3 in /usr/lib/x86_64-linux-gnu/libntl.so.43
-- end stack trace *-

Everything works nicely in Debian unstable, where everything was built against the newer NTL, but this issue is blocking both NTL and and Macaulay2 from migrating to Debian testing. See https://bugs.debian.org/995020.

One band-aid fix would be to just skip the tests that are affected for now until everything migrates and is built against the newer NTL, but I wonder if this is revealing a deeper bug somewhere?