Add compactlowercase and COMPACTUPPERCASE casing rules

[coriolinus]

These rules transform the input case and eliminate spacing. Motivating example: (from serde-rs/serde#2153 (comment)):

I'm consuming an API which has objects with fields firstname, lastname, streetnumber, and the like. In a perfect world, I could write this as

#[derive(Debug, Deserialize, Serialize)]
#[serde(rename_all="lowercase")]
struct FromForeignApi {
    first_name: String,
    last_name: String,
    street_number: String,
    // etc
}

However, attempting this, I encounter the surprising situation that unlike most rename_all rules, lowercase does not adjust the spacing between the words.

[jplatte]

Sorry, but I'm not going to accept this PR with yet another extremely uncommon casing rule. There's a chance custom casing rules will be supported in the future, see #6.

[coriolinus]

This is more uncommon than train case? Looking at #38, the motivation seems to be as simple as "convert_case does it". convert_case handles these also.

[jplatte]

Squashing multi-word identifiers / titles / taglines into one all-lowercase or all-uppercase word is very uncommon I would say. Also (which is probably the distinguishing factor against Train-Case) there is no established name for this casing rule.

[coriolinus]

I probably would have agreed that it's uncommon, until I ran into the problem of having to manually annotate dozens of #[serde(rename="whatever")] attributes in order to support that external API. At that point, it went from "whatever, it's uncommon" to "this is very important for me right now."

It is frustrating that this has been rejected based on a gut instinct about how common this is, plus the absence of an established name. Literally any written policy for contributing a casing rule, including "don't bother unless you've got a contributor's approval ahead of time", would be an improvement on the current status quo.

[coriolinus]

The "established name" heuristic is particularly frustrating, given that as of right now, googling for "train case" produces a page about luggage; there are no links to the casing rule on that front page.

[jplatte]

I can understand that you're not happy about the absence of a written policy for new casing rules, but this just one of many libraries I maintain in my spare, and generally very low on my list of priorities.

However, it also seems like you are seeing some relation between this crate and serde, and I just want to note that serde ships its own case conversion implementation that has nothing to do with heck.