perf: optimize label sanitization

[kruskall]

The following code was used to microbenchmark the sanitizeLabel function:

package modelpb

import "testing"

func BenchmarkSanitize(b *testing.B) {
        small := "foobar"
        l := "foobarfoobarfoobarfoobarfoobarfoobarfoobar"
        b.Run("small", func(b *testing.B) {
                for i := 0; i < b.N; i++ {
                        sanitizeLabelKey(small)
                }
        })
        b.Run("l", func(b *testing.B) {
                for i := 0; i < b.N; i++ {
                        sanitizeLabelKey(l)
                }
        })
}

Benchmarks show a performance improvement for both small and large label keys:

                  │  before.txt  │              after.txt              │
                  │    sec/op    │   sec/op     vs base                │
Sanitize/small-20   16.100n ± 1%   9.191n ± 1%  -42.91% (p=0.000 n=10)
Sanitize/l-20        81.55n ± 1%   29.35n ± 2%  -64.01% (p=0.000 n=10)
geomean              36.23n        16.42n       -54.67%

                  │  before.txt  │              after.txt              │
                  │     B/op     │    B/op     vs base                 │
Sanitize/small-20   0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
Sanitize/l-20       0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
geomean                        ²               +0.00%                ²
¹ all samples are equal
² summaries must be >0 to compute geomean

                  │  before.txt  │              after.txt              │
                  │  allocs/op   │ allocs/op   vs base                 │
Sanitize/small-20   0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹
Sanitize/l-20       0.000 ± 0%     0.000 ± 0%       ~ (p=1.000 n=10) ¹

Closes #49