Add formatting of small floats (10^-n)

[Aras14HD]

This enables formatting small floats like 6.631537423e-34 as 6.631537423 × 10^-34, by allowing negative exponents in format_float (and some more logic to handle that).

Also fixes #315

Resolves #308

[tolik518]

On your test-run factorion replied on 6.631537423e-34! with Factorial of 6.631537423 × 10-34 is approximately 1, before that it was replying with Factorial of 0.000000000000000000000000663154 is approximately 0,999999999999999999999999617217.

https://www.reddit.com/r/testingground4bots/comments/1rabfbh/comment/o77a0e8/

The before reply was more precise

[Aras14HD]

Ok investigating this I found, that the true result is 9.999999999999999999999999999999996172172717063657224823672597116605474397837555828144330523967376628717255627642017055031081750434013261087338778618778539329115394750197229780734699344408893837837414084874090262554352460558789098273872863113806646310596473029774819683214882891955307493169687938970352854185035e-1 not 0.999999999999999999999999617217. That's a lot more 9s in reality (more than 30).

Even though I confirmed we now do better (as more correct), I wanna figure out why it was wrong previously.

[Aras14HD]

Checked on master now, and can't reproduce that response, just get 0.1. What options (consts like float_precision) did you use?

[tolik518]

Checked on master now, and can't reproduce that response, just get 0.1. What options (consts like float_precision) did you use?

I needed a while to reproduce, but it happens to me only if the language is NOT en.

Example-test:

#[test]
fn teststuff() {
    let consts = Consts::default();
    let comment = Comment::new(
        "6.631537423e-34! ",
        (),
        Commands::TERMIAL,
        MAX_LENGTH,
        "de",
    )
        .extract(&consts)
        .calc(&consts);

    let reply = comment.get_reply(&consts);
    assert_eq!(
        reply,
        "Fakultät von 0,000000000000000000000000663154 ist ungefähr 0,999999999999999999999999617217 \n\n\n*^(Dieser Kommentar wurde automatisch geschrieben | [Quelltext](http://f.r0.fyi))*"
    );
}

for en the test would fail, for en the reply would have been Factorial of 0 is approximately 0.1 \n\n\n*^(This action was performed by a bot | [Source code](http://f.r0.fyi))* as you stated.

[Aras14HD]

Oh, It's not doing the same calculation at all. Since the decimal separator is ',' and not '.' the latter is simply ignored.

So on en we calculate the factorial of 6.631537423e-34, but on ru and de, we calculate the factorial of 6631537423e-34.

I am unsure on what should be done. To "fix" this we would have to somehow find, when it is not used as a separator, specifying some pattern of separators per locale (char and interval?) and then? Treat it as a decimal separator or give an error or something?

[tolik518]

Ohh, right! That was my bad!
Its basically 0.0000000000000000000000000000000006631537423 vs 0.0000000000000000000000006631537423.

This is whole other issue as it appears. We would need a heuristic approach that makes assumptions about the separator.

I can merge the PR later in the evening, as it seems to work as intended to me.