Bump dependencies, refactor hash() option

[novemberborn]

Apologies for combining all of this into one PR :) First couple commits update some dev dependencies, add linting and clean up some whitespace. I've also removed unused spies from the tests.

---

The real meat of this PR is replacing the hash option. Both AVA and nyc use this option to ensure some extra data is included in the hash, and to store a mapping from the hash to a file.

Overriding the hash function in order to achieve these goals seems too error-prone, especially since it relies on the consuming code to properly invoke md5-hex. This PR replaces the hash option by a hashData function an an onHash callback:

• hashData receives the original input and additional data and may return a string, buffer or array of either. This is then included in the hash.
• onHash is called with the input, additional data, and the resulting hash. Calling code can use this to maintain mappings.

I've tried this locally with AVA and it works fine. Plus AVA no longer needs a dependency on md5-hex which is nice.

This is a breaking change though!

---

With the above changes it becomes possible to include a package hash for caching-precompiler itself, which should be useful if the storage format is changed. I believe that'll be necessary for istanbuljs/nyc#217, which requires multiple files to be generated for a single input.

[coveralls]

Coverage remained the same at 100.0% when pulling 106f900 on novemberborn:update-and-hash-refactor into 508395e on jamestalmage:master.

[jamestalmage]

newlines before and after if blocks.

I forget to follow this myself half the time. But @sindresorhus is going to introduce it into XO as soon as the rule exists, so might as well start following the rule.

[novemberborn]

Do you mean:

        var data = [ownHash || getOwnHash(), input];

        if (salt) {
            data.push(salt);
        }
        if (hashData) {
            data = data.concat(hashData(input, metadata));
        }

Or:

        var data = [ownHash || getOwnHash(), input];

        if (salt) {
            data.push(salt);
        }

        if (hashData) {
            data = data.concat(hashData(input, metadata));
        }

Cause there's more if-blocks to change in case of the latter. Which makes me thing this should wait until XO complains.

[jamestalmage]

The latter.
We can do a cleanup commit later. That's fine.

[novemberborn]

@jamestalmage force-pushed to bump XO to 0.15.0. Let me know what you want me to do with the if conditions. Have your concerns about package-hash been alleviated?

[coveralls]

Coverage remained the same at 100.0% when pulling c74af0f on novemberborn:update-and-hash-refactor into 508395e on jamestalmage:master.

[jamestalmage]

one more #

[jamestalmage]

A few comments on the docs. if statements are fine as is until we find a better linter rule.

Otherwise, LGTM

[novemberborn]

Force-pushed again. Updated readme (and fixed an existing typo). Bumped package-hash to the freshly published 1.2.0.

[coveralls]

Coverage remained the same at 100.0% when pulling 45d43dc on novemberborn:update-and-hash-refactor into 508395e on jamestalmage:master.