Auto-parama-roundup (WIP)

[simoncozens]

This PR implements #8. The general idea is to add a helper for casting rays from various points and directions, looking at the intersections between the ray and the glyph, and then trying to organise those intersections into pairs a relatively similar distance apart. i.e.

should return 177,177 and ignore the 802 as an outlier.

It also adds some routines for sweeping a ray around a portion of the glyph (which we call "jittering") and then taking the average discarding outliers. This allows for us to measure stems while ignoring irrelevant features. i.e. in this glyph, we should get a value for red (and maybe orange) but not blue:

Using these new routines, we can find values for XOPQ, XOUC, XOLC, XOFI, YOPQ, YOUC, YOLC and YOFI.

The next step is to add routines which "expect" a number of intersections and then choose a value from the set which is not the most common. (i.e. with the same H above, we should say "expect to see four intersections and when you do, give me the pair with the maximum distance between points) this is how we return the 802 to get XTRA.

[simoncozens]

Actually, plan B: I'm going to use winding numbers to differentiate between transparent and ink.

[simoncozens]

Now I am starting to get some interesting results:

However it's not clear to me what the parametric values mean for italic fonts. Should I measure it like this:

or like this:

[simoncozens]

Looking at Amstelvar and Robotoflex, parametric values should not correct for slant.