Update .npmignore due to snappy changes

[ralphtheninja]

See Level/rocksdb#80 (comment)

We get the testdata folder in the tarball now and unpacked it's 2.9Mb.

[filoozom]

Would updating .npmignore actually work for Git Submodules? If npm simply runs git submodule -q update --init --recursive, then I wouldn't count on npm to delete files from the git repo, but I guess I should test that first.

This also means that this package now depends on git, which you initially didn't want if I remember correctly. I thought we pushed the whole source code to npm. Maybe ignoring .gitmodules could work?

And if we actually depend on git now, then we might as well only fetch Submodules if the library needs to be built and can't use prebuild packages.

[ralphtheninja]

Would updating .npmignore actually work for Git Submodules?

Well, primarily it's for publishing to npm so we don't publish the testdata folder.

This also means that this package now depends on git, which you initially didn't want if I remember correctly.

No, we don't depend on git. Why would this be the case?

[filoozom]

Maybe I misunderstood something, but based on the conversation in Level/rocksdb#80 I assumed that npm ran git after install as opposed to on prepublish, which might not be the case. Did you publish a test version on npm with Git Submodules?

[vweevers]

Maybe I misunderstood something, but based on the conversation in Level/rocksdb#80 [..]

That conversation was about npm i Level/leveldown rather than the usual npm i leveldown (which does not need git).