Improves benchmark comparison

[springmeyer]

This fixes a few issues in / improves the benchmark script:

• Option to test compositing from compressed tiles is called --compress but usage said --compressed
• vtcomposite mode was not tracking memory and memory stats were not output at all
• We were using vt.addDataSync for node-mapnik, whereas in production we use the async vt.addData due to https://github.com/mapbox/tilelive-vector/blob/master/backend.js#L113 (which means that gzip decompression is async)
• refactors code to avoid duplication between vtcomposite and node-mapnik modes

/cc @millzpaugh @artemp

[springmeyer]

Looks like the async node-mapnik approach is significantly faster (in some cases):

node bench/bench.js --iterations 500 --concurrency 4 --package node-mapnik --compress
1: single tile in/out ... 2392 runs/s (209ms)
2: two different tiles at the same zoom level, zero buffer ... 484 runs/s (1034ms)
3: two different tiles from different zoom levels (separated by one zoom), zero buffer ... 209 runs/s (2391ms)
4: two different tiles from different zoom levels (separated by more than one zoom), zero buffer ... 310 runs/s (1612ms)
5: tiles completely made of points, overzooming, no properties ... 3788 runs/s (132ms)
6: tiles completely made of points, same zoom, no properties ... 15152 runs/s (33ms)
7: tiles completely made of points, overzoooming, lots of properties ... 1276 runs/s (392ms)
8: tiles completely made of points, same zoom, lots of properties ... 10417 runs/s (48ms)
9: buffer_size 128 - tiles completely made of points, same zoom, lots of properties ... 9804 runs/s (51ms)
10: tiles completely made of linestrings, overzooming and lots of properties ... 480 runs/s (1042ms)
11: tiles completely made of polygons, overzooming and lots of properties ... 161 runs/s (3106ms)
12: tiles completely made of points and linestrings, overzooming and lots of properties ... 3937 runs/s (127ms)
13: returns compressed buffer - tiles completely made of points and linestrings, overzooming and lots of properties ... 3937 runs/s (127ms)
14: buffer_size 128 - tiles completely made of points and linestrings, overzooming and lots of properties ... 4065 runs/s (123ms)
15: tiles completely made of points and linestrings, overzooming (2x) and lots of properties ... 5882 runs/s (85ms)
16: tiles completely made of polygons, overzooming and lots of properties ... 442 runs/s (1132ms)
17: tiles completely made of polygons, overzooming (2x) and lots of properties ... 852 runs/s (587ms)
18: return compressed buffer - tiles completely made of polygons, overzooming (2x) and lots of properties ... 845 runs/s (592ms)
19: buffer_size 4096 - tiles completely made of polygons, overzooming (2x) and lots of properties ... 833 runs/s (600ms)

Against master/sync addData:

$ node bench/bench.js --iterations 500 --concurrency 4 --package node-mapnik --compress

1: single tile in/out ... 1374 runs/s (364ms)
2: two different tiles at the same zoom level, zero buffer ... 237 runs/s (2106ms)
3: two different tiles from different zoom levels (separated by one zoom), zero buffer ... 204 runs/s (2452ms)
4: two different tiles from different zoom levels (separated by more than one zoom), zero buffer ... 239 runs/s (2096ms)
5: tiles completely made of points, overzooming, no properties ... 3968 runs/s (126ms)
6: tiles completely made of points, same zoom, no properties ... 7463 runs/s (67ms)
7: tiles completely made of points, overzoooming, lots of properties ... 1232 runs/s (406ms)
8: tiles completely made of points, same zoom, lots of properties ... 4310 runs/s (116ms)
9: buffer_size 128 - tiles completely made of points, same zoom, lots of properties ... 4167 runs/s (120ms)
10: tiles completely made of linestrings, overzooming and lots of properties ... 469 runs/s (1067ms)
11: tiles completely made of polygons, overzooming and lots of properties ... 157 runs/s (3180ms)
12: tiles completely made of points and linestrings, overzooming and lots of properties ... 4274 runs/s (117ms)
13: returns compressed buffer - tiles completely made of points and linestrings, overzooming and lots of properties ... 3968 runs/s (126ms)
14: buffer_size 128 - tiles completely made of points and linestrings, overzooming and lots of properties ... 4464 runs/s (112ms)
15: tiles completely made of points and linestrings, overzooming (2x) and lots of properties ... 6024 runs/s (83ms)
16: tiles completely made of polygons, overzooming and lots of properties ... 432 runs/s (1158ms)
17: tiles completely made of polygons, overzooming (2x) and lots of properties ... 760 runs/s (658ms)
18: return compressed buffer - tiles completely made of polygons, overzooming (2x) and lots of properties ... 839 runs/s (596ms)
19: buffer_size 4096 - tiles completely made of polygons, overzooming (2x) and lots of properties ... 852 runs/s (587ms)