refactor: use toml.Unmarshal instead of toml.Decode

[alexandear]

The PR refactors parseConfig implementation to use toml.Unmarshal instead of toml.Decode. The first accepts the data as []byte, the second as string. This removes one conversion between from []byte to string.

Benchmark says that this refactoring saves one allocation without increasing parseConfig complexity.

Benchmarks

func BenchmarkParseConfig(b *testing.B) {
	var config lint.Config

	configPath := filepath.Join("..", "full.toml")

	b.ResetTimer()
	for i := 0; i < b.N; i++ {
		err := parseConfig(configPath, &config)
		if err != nil {
			b.Fatalf("failed to parse config: %v", err)
		}
	}
}

func BenchmarkParseConfigString(b *testing.B) {
	var config lint.Config

	configPath := filepath.Join("..", "full.toml")

	b.ResetTimer()
	for i := 0; i < b.N; i++ {
		err := parseConfigString(configPath, &config)
		if err != nil {
			b.Fatalf("failed to parse config: %v", err)
		}
	}
}

❯ go test -benchmem -bench=. ./...
goos: darwin
goarch: arm64
pkg: github.com/mgechev/revive/config
cpu: Apple M1 Pro
BenchmarkParseConfig-8              5030            213835 ns/op          158632 B/op       1789 allocs/op
BenchmarkParseConfigString-8        5779            210645 ns/op          161789 B/op       1790 allocs/op
PASS
ok      github.com/mgechev/revive/config        2.752s