Fix bug where close-to-collinear line segments with thickness >= 2 got size `f64::INFINITTY as i32` as coordinates.

[el-hult]

Bugfix for #562 .

The root cause of the bug was a comparison of floats, but to make sure I got it right, I changed the code away from early returns (that was part of why the bug surfaced) and added some more comments.

This script shows example thick lines with inside and outside corners, and the capping feature.

use plotters::prelude::*;
pub fn main() -> Result<(), Box<dyn std::error::Error>> {
    let root = BitMapBackend::new("test.bmp", (1000, 1000)).into_drawing_area();
    let mut chart = ChartBuilder::on(&root)
        .build_cartesian_2d(0.0..1000.0, 0.0..1000.0)?;
    let pahts = vec![
        PathElement::new(
            vec![(336.0, 614.0), (339.0, 674.0),(341.0,714.0)],
            ShapeStyle::from(RED).stroke_width(2),
        ),
        PathElement::new(
            vec![(100.0, 100.0), (150.0, 150.0),(200.0,100.0)],
            ShapeStyle::from(BLUE).stroke_width(2),
        ),
        PathElement::new(
            vec![(400.0, 400.0), (400.0, 450.0),(400.0,500.0)],
            ShapeStyle::from(GREEN).stroke_width(5),
        ),
        PathElement::new(
            vec![(900.0, 410.0), (600.0, 400.0),(900.0,500.0), (1000.0, 1000.0)],
            ShapeStyle::from(YELLOW).stroke_width(10),
        ),
    ];
    chart.draw_series(pahts.into_iter())?;
    Ok(())
}

[el-hult]

Hi @AaronErhardt You asked in #345 that we should ping you if the issue goes without notice. Time to take a look at this PR?

I have done what I hope is a quite minimal fix. I have created a unit test and an example that shows the output. Is okay?

[el-hult]

@AaronErhardt I did revert a bit of the changes, to make the PR smaller and easier to review. Looks okay now?

[el-hult]

Now I can go back to use the main plotters-rs instead of my fork. Thanks a lot!