Implement target type inference for embedding mode

[alilleybrinker]

In order to support embedding mode for manifest creation, we need to able to infer the "target type," the type of artifact for which the manifest is being created. This target type informs how the embedding should be performed.

In particular, today the TargetType enum looks like this:

#[derive(Debug)]
enum TargetType {
    KnownBinaryType(BinaryType),
    KnownTextType(TextType),
    Unknown,
}

#[derive(Debug)]
enum BinaryType {
    ElfFile,
}

#[derive(Debug)]
enum TextType {
    PrefixComments { prefix: String },
    WrappedComments { prefix: String, suffix: String },
}

As shown here, that means there's a need to support inference for three different types of targets:

• ELF binaries
• Text formats with prefix based comments
• Text formats with wrapped comments

Of course, there are an enormous number of possible text formats that would support prefix comments or wrapped comments. For this we'll likely want to rely first on third-party projects which exist already and infer the format used for existing text files, things like GitHub's Linguist project (though that's in Ruby, I am unsure of what conceptually-equivalent things exist in Rust today), and then map any languages identified in that list to indicate they use prefix comments or wrapped comments as appropriate.