Webpack next per-source minification

[guybedford]

This implements #115 on top of #162, handling minification per-source, and therefore also caching minification per source.

The previous results it turns out were actually without minification due to #168. Here are the numbers with minification:

The build time summary of the self build here is:

• With cache, with minification before this PR: First build 64s, Subsequent builds 47s
• With cache, with minification after this PR: First build 64s, Subsequent builds 36s

[codecov-io]

Codecov Report

Merging #169 into webpack-next-cache will increase coverage by 0.01%.
The diff coverage is 90%.

@@                  Coverage Diff                   @@
##           webpack-next-cache     #169      +/-   ##
======================================================
+ Coverage               73.32%   73.33%   +0.01%     
======================================================
  Files                      12       13       +1     
  Lines                     536      540       +4     
======================================================
+ Hits                      393      396       +3     
- Misses                    143      144       +1

Impacted Files	Coverage Δ
src/index.js	93.25% <ø> (-0.43%)	⬇️
src/loaders/terser-loader.js	90% <90%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update cf5cd9d...d6961cf. Read the comment docs.

[guybedford]

Here's the difference in footprint on the self-build:

• 2,691,197 bytes with per-source minification
• 2,456,293 bytes with whole-file minification

the difference there is basically the wrappers not being minified - function (module, exports, __webpack_require__) sort of thing instead of function(a,b,c). There's no easy safe way to hack in the minification of the wrappers though unfortunately.

So the question then is the extra 10s on the self build (plus inability to parallelize) worth 100KB? I'd argue it might not be...

[rauchg]

@guybedford I agree it's not. Keep in mind that once we investigate #51 I'm not even sure minification is worth it to begin with