New output format: Seven-bit coding (-F7)

[lifthrasiir]

This is an approach used by base-122 and optimal if you are using UTF-8. It recognizes the forbidden code units (in this case, \r, $, `, \ and possibly \0) and encodes it and the following code unit into two-byte UTF-8 sequences:

0xxxxxxx                for allowed code units xx
110xxxyy 10yyyyyz       for forbidden code units xx (indexed) and following code unit yy

It should be noted that two-byte UTF-8 sequences have up to two spare bits to spare.

• The index to forbidden code units are encoded to 3 bits (the first index 0 should be reserved to make it valid UTF-8), so we definitely have a single unused bit (denoted as z above).
• Not all indices to forbidden code units would be used. Since coding into two-byte sequence actually saves the space, we can assign otherwise allowed code units into these indices.

We may need a change to ANSEncoder to leverage this spare bit, so it is not as trivial.