use a simpler hashing algorithm when a file system is available

[quantizor]

fixes #105

By not including the file contents in the hash, the generated IDs
should be more durable and less susceptible to minor differences
in file content due to a babel plugin being applied to ssr vs
client config or something like that

[quantizor]

Going to add a snapshot serializer to do the whitespace removal stuff done in
8037d30

[mxstbr]

This looks good in terms of the code that changed, but the snapshot change is making it very hard to review what exactly this affects/if it works. 😕

[quantizor]

@mxstbr even with the separate commits? I'll merge the whitespace serializer back into the first two commits to make it as clear as possible

[quantizor]

@mxstbr I'll split up 8037d30 such that the rename from before.js to index.js happens in a separate commit. That way you'll be able to verify that the original snapshots are accurately recreated.

I ended up dropping the rename from before.js to index.js so now in 2445cca it should be clear that the hashes did not change

[quantizor]

There we go.

[mxstbr]

LGTM, @philpl? Any reasons why we shouldn't do it this way?

[mxstbr]

Shipping this and the ref styled-components PR! styled-components/styled-components#1381