bug(ansi): Truncate, TruncateLeft etc do not obey length in many cases

### Describe the bug

In many cases, `Truncate`, `TruncateLeft`, and potentially other functions do not obey the provided length, at least lengths that get returned by `StringWidth`.

### To Reproduce

Use the following patch to add fuzzing for `Truncate` and `TruncateLeft`:

```patch
diff --git a/ansi/truncate_test.go b/ansi/truncate_test.go
index decf11a..17e5ba3 100644
--- a/ansi/truncate_test.go
+++ b/ansi/truncate_test.go
@@ -1,6 +1,7 @@
package ansi

import (
+ "math/rand"
"testing"
)

@@ -458,3 +459,35 @@ func TestByteToGraphemeRange(t *testing.T) {
})
}
}
+
+func FuzzTruncate(f *testing.F) {
+ for _, tc := range tcases {
+ f.Add(tc.input)
+ }
+
+ f.Fuzz(func(t *testing.T, input string) {
+ n := rand.Intn(100) + 1
+ result := Truncate(input, n, "…")
+ w := StringWidth(result)
+
+ if w > n {
+ t.Errorf("expected width to be no greater than %d, got %d", n, w)
+ }
+ })
+}
+
+func FuzzTruncateLeft(f *testing.F) {
+ for _, tc := range tcases {
+ f.Add(tc.input)
+ }
+
+ f.Fuzz(func(t *testing.T, input string) {
+ n := rand.Intn(100) + 1
+ result := TruncateLeft(input, n, "…")
+ w := StringWidth(result)
+
+ if w > n {
+ t.Errorf("expected width to be no greater than %d, got %d", n, w)
+ }
+ })
+}
```

And run with:

```console
$ go test -run=".*" -fuzz="^FuzzTruncate$" ./ansi/
# or
$ go test -run=".*" -fuzz="^FuzzTruncateLeft$" ./ansi/
```

I get failures nearly instantly. An example of 1:

```
string("\x9f \xec؋\xf00")
```

### Expected behavior

The above fuzzing methods simply check to see if the resulting output from either function is ever **longer** (according to `ansi.StringWidth()`) than the provided max length passed to the associated function. I would expect that the result to always be truncated and **never exceed the length provided as the input**, at least according to the "length" that `ansi.StringWidth()` also calculates.

Start in ansi/truncate_test.go and run the provided Go fuzz commands for FuzzTruncate and FuzzTruncateLeft to reproduce the width failures. Trace Truncate and TruncateLeft from those tests, then verify that fuzz-generated results never have StringWidth greater than the supplied length for either function.

Written by the indexing model from the issue text.