perf(slogpretty): precompute group prefix and avoid value allocations

[tigerwill90]

Some small optimisation on the builtin slog handler to reduce memory allocation.

goos: darwin
goarch: arm64
pkg: github.com/fox-toolkit/fox/internal/slogpretty
cpu: Apple M4 Max
                           │ /tmp/slog_before.txt │         /tmp/slog_after.txt         │
                           │        sec/op        │   sec/op     vs base                │
Handle-16                             234.4n ± 1%   159.8n ± 1%  -31.81% (p=0.000 n=10)
HandleWithGroup-16                    271.6n ± 1%   164.6n ± 1%  -39.40% (p=0.000 n=10)
HandleWithAttrsAndGroup-16            370.7n ± 1%   210.5n ± 1%  -43.22% (p=0.000 n=10)
geomean                               286.8n        176.9n       -38.32%

[codecov]

Codecov Report

❌ Patch coverage is 43.54839% with 35 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
internal/slogpretty/handler.go	43.54%	33 Missing and 2 partials ⚠️

📢 Thoughts on this report? Let us know!