"generatepath lb" slices timestamps with colon

Nobody has claimed this yet.

When running `generatepath lb` on a URL with a number of `:`'s, everything beyond the second-to-last semicolon is removed from the URL, resulting in unexpected URLs returned.

Example URL input (from within a CloudFront log file)

```
GET http://api.example.com:80/v4/mapsam.semicolon-test/2/3/1.vector.pbf?access_token=pk.token&style=mapbox://styles/mapsam/1234@2017-05-31T22:05:32.562Z
```

Results in the following output

```
/v4/mapsam.semicolon-test/2/3/1.vector.pbf?access_token=pk.token&style=mapbox://styles/mapsam/1234@2017-05-31T22
```

It appears it's just related to time stamps. The following URL with a number of semi colons works just fine:

Input
```
GET http://api.example.com:80/v1/thing/my.id?access_token=hello:world&another=thing:with:a:semicolon/and/some//slashes:waka
```

Output
```
/v1/thing/my.id?access_token=hello:world&another=thing:with:a:semicolon/and/some//slashes:waka
```

But adding just a timestamp results in it getting cut off:

Input
```
GET http://api.example.com:80/v1/thing/my.id?time=2017-05-31T22:05:32.562Z
```

Output
```
/v1/thing/my.id?time=2017-05-31T22
```

cc @jakepruitt @springmeyer

Start by tracing the `generatepath lb` command and its URL parsing entry point, then reproduce the timestamp example from the issue. Done means the full `2017-05-31T22:05:32.562Z` query value and the rest of the URL are preserved, while the existing colon-containing example continues to work.

Written by the indexing model from the issue text.