Suggest using format precision for floats instead of casting to integer

What it does

What does this lint do?
Linting code where a floating point value (f64/f32) is casted to an integer just for printing/formatting. The intention behind that is to remove the fraction part of that number and only display the integer itself. This is closely related to cast_possible_truncation.

This kind of lint could be applied to everything, that takes a fmt string, e.g. format, panic, write, ... .

Categories (optional)

clippy::pedantic maybe clippy::complexity or clippy::suspicious. I'm not sure here (:

What is the advantage of the recommended code over the original code

If the value should be NaN or INFINITY, it will be printed/formatted as such and easier to detect instead of silently converting it to 0 or min/max of the type.

Drawbacks

Could be a false positive.

Example

fn print_it(a: f64) {
    println!("{}", a as i64);
}

Could be written as:

fn print_it(a: f64) {
    println!("{:.0}", a);
}

Start from the issue's Rust examples and compare the proposed formatting precision with the existing cast-related lint behavior. Determine which formatting macros and float-to-integer casts are in scope, then define tests covering normal values, NaN, infinity, and possible false positives. Done means the lint scope, category, diagnostics, and test expectations are agreed.