FileType guesstimating needs refactoring

[jtmoon79]

Summary

file type estimating (guessing) is kind of messy.

Current behavior

1. the use of Mimetype adds nearly zero benefit for a lot of code
2. the resultant MimeGuess and FileType is confusing; which one matters when?
3. FileType guessing is hacky name matching

Suggested behavior

1. remove Mimetype and MimeGuess entirely (affects BlockReader should receive MimeGuess #15)
2. more robust and systematic approach to determining FileType based on the file name

If 1. and 2. are completed then a new Issue should be created around allowing the filepreprocessor.rs to read the zero block of the file and do some kind of magic fingerprint matching as well.
That change leads to another very large change wherein multiple FileTypes may be returned during file preprocessing, where the appropriate Reader is attempted and if it fails then the next Reader is attempted.

[jtmoon79]

new Issue should be created around allowing the filepreprocessor.rs to read the zero block of the file and do some kind of magic fingerprint matching as well.

• https://github.com/bojand/infer for determining file type
• https://github.com/anemele/filetype.rs for determining file type
• https://lib.rs/crates/file-format for determining file type
• https://lib.rs/crates/tree_magic_mini for determining file type

For Issue #16

• https://lib.rs/crates/encoding_rs for determining text encoding