Optimizations to the dictionary comparison strategy

[ESultanik]

Take these two JSON files as an example:

$ cat f1.json
{
    "foo": [1, 2, 3],
    "oof": [1, "two", 3]
}
$ cat f2.json
{
    "bar": [1, 2, 3],
    "foo": [1, "two", 3] 
}

By default Graphtage used to try all possible matchings between dictionary key/value pairs; comparing graphtage f1.json f2.json would result in the "foo" key being replaced by "bar" and the "f" in the "oof" key being moved to the front of the string.

This sort of matching is polynomial time in the size of the input, but often is still intractable for large files. Therefore, Graphtage had an option, --no-key-edits or -k that would prevent two dictionary key/value pairs from being compared to each other unless their keys were identical. graphtage -k f1.json f2.json would have resulted in the 2 being replaced by "two", the entire "oof" key/value pair being removed, and the entire "foo" key/value pair being added.

This PR…

1. generalizes these two options with a new --dict-strategy/-ds option which sets the strategy: match for the old default behavior and none for the old --no-key-edits behavior. The --no-key-edits option still exists, but now is an alias to --dict-strategy none.
2. adds a new strategy, --dict-strategy auto, which is now the default, that behaves exactly the same as the match strategy, but in the event that two key/value pairs have then exact same key, then they are automatically matched.

graphtage --dict-strategy auto f1.json f2.json will now result in 2 being replaces with "two", oof being replaced by bar, and "two" being replaced by 2.