Improve the cognitive_complexity lint to show what lines are causing the complexity

We could improve the cognitive_complexity lint further by showing the ^ sign under tokens which cause the increment of cognitive complexity. Doing so would let users know what actually was causing the complexity and so they would spend their time more effectively fixing the issue.

fn f(a: u8) {
    if a == 0 {
    } else if a == 3 {
    } else {
    }
}

warning: the function has a cognitive complexity of (3/2)
  --> src/main.rs:12:1
   |
12 | / fn f(a: u8) {
13 | |     if a == 0 {
14 | |     } else if a == 3 {
15 | |     } else {
16 | |     }
17 | | }
   | |_^
   |
   = note: `#[warn(clippy::cognitive_complexity)]` on by default
   = help: you could split it up into multiple smaller functions
   = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#cognitive_complexity

note: these places cause additional cognitive complexity
   |
1 | /     fn f(a: u8) {
2 | |         if a == 0 {
  | |_________^ increments to 1

   |
1 | /     fn f(a: u8) {
2 | |         if a == 0 {
3 | |         } else if a == 3 {
  | |___________^ increments to 2

   |
1 | /     fn f(a: u8) {
2 | |         if a == 0 {
... |
4 | |         } else {
  | |___________^ increments to 3

Right now, for people unfamiliar with the cognitive complexity, it is impossible to know for sure what is causing problems in their code, they have to study the paper. Also, even people who had it known before the problem appeared, could simply have forgotten it. This explicit show will not only help them understand what causes a problem here but also will teach them.

Start by locating the Rust implementation and tests for the cognitive_complexity lint, then reproduce the issue with the example in the description. The work is done when diagnostics identify the tokens that add complexity with caret markers and increment values, while retaining the existing warning output.