[DRAFT] [starvation] treat pthread mutexes as non-recursive

[VladimirMakaev]

Stack: #2182 → #2210 (review only the top commit; lands after #2182)

Starvation ignores re-acquisitions of recursive locks, but had no model for pthread_mutex_t and
looked up only C++ classes, so every pthread mutex counted as recursive and self deadlocks on
them went unreported.

struct Device { pthread_mutex_t mutex; };
void reset(struct Device* d) {
  pthread_mutex_lock(&d->mutex);
  pthread_mutex_unlock(&d->mutex);
}
void close_device(struct Device* d) {
  pthread_mutex_lock(&d->mutex);
  reset(d); // Infer missed: self deadlock
  pthread_mutex_unlock(&d->mutex);
}

Lock models now cover C structs and unions, pthread_mutex_t is non-recursive, boost's recursive
mutexes are modelled as recursive, and "recursive" in --lock-model overrides the built-in
setting. Global locks are matched through their extra pointer layer; file-static globals are
distinct per translation unit. Array elements (eg lock stripes) count as recursive, as indices are
not tracked, and recursive relocks that appear only once a callee's locks are substituted are
ignored too.

Limitations:

• recursive pthread mutexes are not recognised; a wrapper class needs a --lock-model listing its
  methods with "recursive": true.
• relocking the same array element is missed.

Test plan

New tests in c/starvation (self_deadlock.c, static_lock_{1,2}.c) and cpp/starvation (pthread.cpp,
recursive.cpp, custom_guards.cpp), with FP_/FN_ tests for the limitations. The codetoanalyze
tests pass.