Add lock to avoid segfault while resizing

[dylanrb123]

There is some unsafe memory access in the resizeIndex method, where it deletes and moves around memory while other threads may be accessing it. This PR adds a read-write lock to ensure that the resizeIndex function has exclusive access to the underlying resources when it needs to move the memory around

[psobot]

Do we also need to take a shared lock over resizeLock when calling addPoint? (addPoint is reading the data structures that get mutated while calling resizeIndex, and IIRC would also hit a segfault if we call addItems in parallel with resizeIndex.)

[dylanrb123]

Do we also need to take a shared lock over resizeLock when calling addPoint? (addPoint is reading the data structures that get mutated while calling resizeIndex, and IIRC would also hit a segfault if we call addItems in parallel with resizeIndex.)

addPoint only gets called after resizeIndex returns in the addItems function of TypedIndex so that wouldn't be possible in normal operation, however if someone is manually calling resizeIndex in one thread while calling addItems in another then it could cause trouble. I could take the shared lock just in case but seems pretty unlikely to happen in practice, what do you think?

[psobot]

I think it's still definitely worth locking there too - there are cases where multiple concurrent calls to addItems could still cause a segfault here; i.e.:

Thread A	Thread B
Call addItems
	Call addItems
Call addPoint
cur_element_count++
	Index full, call resizeIndex
	data_level0_memory_new = realloc(data_level0_memory_, ...)
*data_level0_memory_ 💥

[dylanrb123]

👍 sounds good, will try to replicate in test and add the lock