pybind: Add S1Interval bindings (#524)

[deustis]

Add pybind11 bindings for S1Interval.

Also add interface notes to the python README

(Part of a series of addressing #524)

[deustis]

@jmr, next installment here. I've added some updates to the README that explain some of the user-facing interface choices for the python bindings. That might be a good place to start for the review. Feel free to suggest any changes.

[deustis]

@jmr, thanks for the detailed comments. I think once we iron out some of these interface choices it will be smooth sailing. Definitely worth spending the time up-front to ensure consistency/correctness.

I've addressed most of your comments. The main open question is about how to handle the debug assertions. Let me know your thoughts there and I'll make those changes.

[jmr]

Is there a reason you picked S1Interval? Seems obscure and low level.

[deustis]

Hi there, just a quick note -- I was traveling last week and working against some deadlines this week but still intending to circle back on this!

Is there a reason you picked S1Interval? Seems obscure and low level.

No particular reason... starting with the lowest level primitives and working my way up to S2Cell.

[deustis]

@jmr, apologies for the delay. I'm back and have updated the the PR as discussed.

The python interface will now raise value errors for invalid inputs to avoid any possibility of triggering the DCHECKS. In the case of S1Angle this basically means ensuring all of the points are within +/- pi. I've added a static helper to the C++ code to make this easier so we aren't having to do any math in the bindings themselves.

[deustis]

@jmr, friendly ping on this PR? I believe I've addressed your earlier comments.

[jmr]

I will take a closer look at this tomorrow.