Much better polygon cover algorithm

[mourner]

Makes polygon cover algorithm pretty much as fast and simple as it can get. Closes #58.

• ~20-25% faster according to benchmarks (update: up to 3x faster)
• fixes cover bugs (see unnecessary tiles removed in the out files)
• much much simpler, tons of code removed

@morganherlocker @yhahn you'll like this!

[mourner]

41 additions and 681 deletions, isn't that just cool? My favourite type of PRs.

[yhahn]

@mourner awesome, RIP local min/max ascii art diagrams :joy:

[morganherlocker]

reviewing...

[mourner]

Pushed 2 more changes:

• faster merge algorithm that leads to ~20% faster polygon cover when it's merged
• fix multipoint not being merged

[mourner]

Pushed another perf improvement that leads to up to 3x gain in some polygon cases (where there are a lot of fills):

# before
scan russia polygon - z6 - z6 x 4,849 ops/sec ±2.24% (23 runs sampled)
scan russia polygon - z8 - z8 x 429 ops/sec ±3.56% (22 runs sampled)
scan russia polygon - z10 - z10 x 25.72 ops/sec ±1.72% (13 runs sampled)
scan russia polygon multizoom - z0 - z9 x 80.18 ops/sec ±2.82% (18 runs sampled)

# after
scan russia polygon - z6 - z6 x 6,493 ops/sec ±2.12% (23 runs sampled)
scan russia polygon - z8 - z8 x 1,001 ops/sec ±2.20% (23 runs sampled)
scan russia polygon - z10 - z10 x 81.71 ops/sec ±8.53% (18 runs sampled)
scan russia polygon multizoom - z0 - z9 x 113 ops/sec ±3.30% (19 runs sampled)

[morganherlocker]

:racehorse: published in v3.0.1!

[mourner]

Ha, you're a risky guy for publishing a patch version after half of the code rewritten from scratch :) Thanks Morgan!

[morganherlocker]

@mourner I thought about doing a major, but decided a patch made more sense. This did not change anything functionally (that we know of), other than bugs. If we find any functional - but valid - differences, we can roll this back and publish a major version bump.

That's semver for you ¯\/(°_o)\/¯