Ported to ES6

[SKalt]

I used rbush and flatbush as an example. I

• adopted eslint with eslint-config-mourner
• added rollup
• swapped in let, const, and import
• added a module field in package.json
• added npm/yarn scripts: perf, pretest, test

Tests pass

eslint index.js test/test.js
tape -r esm test/test.js | faucet
# ✓ simplifies points correctly with the given tolerance
# ✓ just return the points if it has only one point
# ✓ just return the points if it has no points
# # tests 3
# # pass  3
# ✓ ok
# Done in 1.07s.

Performance comparison (no significant difference)

All using node 12.

@v1.2.3

Setup script

```sh git checkout v1.2.3 ```

node ./bench/bench.js
# Benchmarking simplify on 1118 points...
# simplify (HQ) x 9,448 ops/sec ±1.99% (88 runs sampled)
# simplify x 15,110 ops/sec ±2.73% (88 runs sampled)

teardown script

```sh git checkout - ```

Using node + ESM

node -r esm bench/bench.js
# Benchmarking simplify on 1118 points...
# simplify (HQ) x 185 ops/sec ±2.86% (78 runs sampled)
# simplify x 2,144 ops/sec ±3.00% (85 runs sampled)
# Done in 11.14s.

Using current compiled UMD

Setup script

git checkout v1.2.3 -- bench/bench.js

node ./bench/bench.js
# Benchmarking simplify on 1118 points...
# simplify (HQ) x 9,339 ops/sec ±2.03% (90 runs sampled)
# simplify x 14,789 ops/sec ±2.50% (86 runs sampled)

teardown script

```sh git reset --hard ```