[pdq] Constant initialize DCT buffer

[ianwal]

Summary

Changed DCT buffer to be constant initialized.

The function which initializes and returns the DCT static buffer was updated to be reentrant in #1338 in pdqhashing.cpp. The change was to protect filling the static buffer from multiple threads by using std::call_once. However, this is unnecessary because it can just be constant initialized because the buffer is never modified after initialization.

In this PR, the buffer is changed to a std::array, so it can be returned from the function. The convenience pointers to its offsets are changed to references in fillFloatLumaFromRGB.

All of the changes are supported by C++11, which is the default for pdq.

Other misc changes:

• Updated some variable names in the function for readability.

• Added venv/ to .gitignore for dev convenience

References:

cppreference.com initialization article (I don't own the standard):

All non-local variables with static storage duration are initialized as part of program startup, before the execution of the main function begins (unless deferred, see below). All non-local variables with thread-local storage duration are initialized as part of thread launch, sequenced-before the execution of the thread function begins. For both of these classes of variables, initialization occurs in two distinct stages:

Anecdote: If compiled with C++20 (ex. https://godbolt.org/z/jejqn6M4Y), this function actually turns into a lookup table. I thought that was interesting.

Test Plan

Tested vpdq and pdq. All tests pass. No concurrency issues when ran with sanitizers (vpdq CI does this).

[Dcallies]

Not surprised it turns into a lookup table - I had something very similar to this in a previous project that the just made into an array filled with constant numbers.

[Dcallies]

For fun, I found this paper and tried to figure out what this was referring to, and I can't find anything on the page that seems to match.

[Dcallies]

Any ideas for in the long term how we can ensure pdq and vpdq say in sync? Should we write a test that checks that file contents of the vpdq directory match the cpp one?