move CSS build to webpack

[silverwind]

• added new 'make webpack' target
• deprecated 'make js' and 'make css'
• extend webpack config to load the less files
• updated docs

I had to rename the source file of arc-green.less to avoid generating a useless JS entrypoint via webpack-fix-style-only-entries which would not work with different source/destination filenames. I hear that there should be cleaner solutions possible once we upgrade to Webpack 5.

[silverwind]

Partially unrelated, but I reworked the "Building Gitea" in CONTRIBUTING.md to just link to the hacking guide as info in it was partially outdated.

[codecov-io]

Codecov Report

Merging #9983 into master will decrease coverage by <.01%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master    #9983      +/-   ##
==========================================
- Coverage   42.25%   42.25%   -0.01%     
==========================================
  Files         610      610              
  Lines       80387    80387              
==========================================
- Hits        33966    33964       -2     
- Misses      42240    42243       +3     
+ Partials     4181     4180       -1

Impacted Files	Coverage Δ
services/pull/temp_repo.go	31.62% <0%> (-2.57%)	⬇️
models/unit.go	37.03% <0%> (-2.47%)	⬇️
services/pull/patch.go	62.89% <0%> (-1.89%)	⬇️
services/pull/check.go	56.64% <0%> (+2.09%)	⬆️
modules/process/manager.go	78.31% <0%> (+3.61%)	⬆️

---

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 cbceca4...e8e3d4d. Read the comment docs.

[silverwind]

Rebased and squashed.

[zeripath]

@silverwind is there any way we can stop having to rebuild fomantic and the webpack every time we do make build?

My machine is FAST but it's taking a considerable amount of time to test one line changes because of this.

(We could also do with something similar for go generate - it's a real shame that bindata is regenerated despite no changes - I might take a look at that...)

[silverwind]

@zeripath I thought 8ca0730 fixed those useless rebuilds. Try running make --trace build to see why it's building those particular targets.

[silverwind]

Ah, I think I found something that triggers useless fomantic rebuilds, testing..

[silverwind]

#9999 for possible workaround.

Regarding generate: Yes, I'd also prefer it to run on-demand but @guillep2k had some concerns about it last time I tried that, see #9114 (comment).

[zeripath]

I was thinking of doing it in the generate functions themselves... They're the thing that is doing the work.

[silverwind]

@zeripath possible, but I think make is ideal for timestamp checks like that.

Rebased here and also added a touch node_modules in case npm fails to update that for whatever reason.

[silverwind]

Rebased