Avoid detecting other files as SVG

[fcharlie]

eg:

var svgText=`<svg></svg>`

before: image/svg+xml
after: text/plain; charset=utf-8

[gabriel-vasile]

@fcharlie, the benchmarks report that that svg detection time increased from 200ns to 5000ns and memory allocs from 0 to 5. svg is now the slowest detector.

I'm not against an implementation using regex, like the one in this PR, but I'd like to explore other options as well:

• do what the linux file utility does: https://github.com/file/file/blob/5b004b5d76534a0969a50f6e4cf1be9a1e6ef87f/magic/Magdir/sgml#L5-L16 (it's not 100% accurate, <svg></svg> is still reported as svg)
• use golang encoding/xml to parse tokens from the input and check for <svg> token (this one is the most accurate but it might be slow: some benchmarks would help)

Edit: the slowness is not a blocker, as long as detection accuracy is improved. Finding out which approach provides best accuracy/performance trade-off requires more effort and I would appreciate a lot if you helped with that.

[fcharlie]

@gabriel-vasile Introducing regular expressions to detect svg may indeed reduce performance.

Below is a simple svg detection function I implemented with reference to file/file, but I cannot guarantee that it can fully cover the detection of svg files:

var (
	xmlPrefix         = []byte("<?xml version=")
	xmlVersionRegex   = regexp.MustCompile(`['"\ \t]*[0-9.]+['"\ \t]*`)
	svgDocumentPrefix = []byte("<!DOCTYPE svg PUBLIC")
	svgPrefix         = []byte("<svg")
)

func detectSVG(b []byte) bool {
	if bytes.HasPrefix(b, xmlPrefix) {
		mlen := min(len(b), 10+len(xmlPrefix))
		return xmlVersionRegex.Match(b[len(xmlPrefix):mlen]) && bytes.Contains(b, []byte("<svg"))
	}
	if bytes.HasPrefix(b, svgDocumentPrefix) {
		return true
	}
	return bytes.HasPrefix(b, svgPrefix)
}