Draft PR for rewriting of remaining EKFSym functions in C++

[wolterhv]

Created according to commaai/openpilot#2386 (comment)

This is a draft PR to provide visibility into my WIP, to avoid rework.

[wolterhv]

@pd0wm I've finished a first pass on the implementation of the functions for this draft PR, but I'm not sure what tests you referred to in your comment (commaai/openpilot#2386 (comment)). Are you referring to the tests for locationd from the openpilot repo?

[wolterhv]

The chi2_lookup function could be optimised further by

1. skipping the reading of lookup table file header and metadata, in the way assuming that lookup table file is correct and that numbers are stored as doubles and
2. just fetching the boundary values needed for the interpolation instead of the whole row.

The current python implementation relies on the __file__ dynamic attribute to determine the location of the lookup table file. Compiled applications have naturally no knowledge of the location of their source files, so probably a fixed path will have to replace this in the C++ implementation. @pd0wm what is your opinion on this?

[pd0wm]

There really is no need to parse the npy file. Just generate the table as plaintext and include it in a header file.