Replace adm-zip with yauzl

[tamird]

All adm-zip versions after 0.5.10 seem to have bugs:

• 0.5.11 causes tests to fail.
• 0.5.12 causes tests to hang.
• 0.5.13 causes extracted files to be empty.

This also avoids extracting anything other than the clangd binary.

This is an alternative to #34.

[HighCommander4]

This also avoids extracting anything other than the clangd binary.

I don't think we can do that; the clangd binary requires the other files in the zip file (dynamically linked library files, and in most configurations built-in header files) to operate correctly.

Regarding yauzl as a choice of library to use: it seems like it has a lower-level API which is more involved to use. What do you think about going back to unzipper? We switched away from it in #25 due to a bug in it, but it looks like that bug has since been resolved. Switching back to it could potentially be as simple as reverting #25.

also cc @fannheyward for any thoughts

[tamird]

unzipper's issues and pull request count don't inspire confidence :(

https://github.com/ZJONSSON/node-unzipper#readme

I can change the implementation to extract all files, that's straightforward.

[tamird]

@HighCommander4 I've updated this to extract the entire archive, as before.

[HighCommander4]

I'm hesitant about replacing a use of a library that was basically a single function call, with one that involves this much more complexity, with four levels of nested callbacks and such; it seems like this would require more effort to maintain going forward.

I see that #34 has been updated with a change that makes the test pass -- any reason not to go forward with that?

(Also added @hokein for a second opinion.)

[tamird]

The change that was made in #34 makes me even more suspicious. It seems to me that the maintainers of adm-zip are a bit sloppy, and that's enough to make me think we should look for an alternative, and yauzl is the most popular library I could find that also has a reasonable support record.

Having said that, downloading clangd is a relatively small part of this library's functionality, so I don't feel very strongly about this.

If we go with #34 we should document the reason for avoiding upgrading adm-zip, particularly since our tests fail to detect the 0.5.13+ issue.