Don't consider .mdiff or .sdiff files in discoverTestData

[egibs]

Addresses the bug seen here: https://github.com/chainguard-dev/malcontent/pull/636/files#r1845470148

I was considering .mdiff and .sdiff files as part of the code that replaced this Bash:

for f in $(find "${test_data}" -name "*.simple"); do
	prog=$(echo $f | sed -e s#"${test_data}/"## -e s#\.simple\$##)
	if [[ -s "${prog}" ]]; then
		addq ${malcontent} --format=simple --min-file-risk=1 --min-risk=1 --ignore-tags=harmless -o "${f}" analyze "${prog}"
	else
		"*** ${prog} sample is empty or does not exist"
		exit 1
	fi
done
for f in $(find "${test_data}" -name "*.md"); do
	prog=$(echo $f | sed -e s#"${test_data}/"## -e s#\.md\$##)
	if [[ -s "${prog}" ]]; then
		addq ${malcontent} --format=markdown --min-file-risk=1 --min-risk=1 --ignore-tags=harmless -o "${f}" analyze "${prog}"
	else
		"*** ${prog} sample is empty or does not exist"
		exit 1
	fi
done
for f in $(find "${test_data}" -name "*.json"); do
	prog=$(echo $f | sed -e s#"${test_data}/"## -e s#\.json\$##)
	if [[ -s "${prog}" ]]; then
		addq ${malcontent} --format=json --min-file-risk=1 --min-risk=1 -o "${f}" analyze "${prog}"
	else
		"*** ${prog} sample is empty or does not exist"
		exit 1
	fi
done

.mdiff and .sdiff files should only be handled by the diffRefresh function and having them in two places caused drift/clobbering in cases where the filenames without extensions were the same (e.g., ls.json being scanned instead of the ls binary).

With this code, I ran about 10 refresh loops and the problematic ls.mdiff file retained its header consistently:

## Changed: macOS/clean/ls [🟡 MEDIUM → 🔵 LOW]

whereas the original code would flip back and forth between this and ## Changed: macOS/clean/ls.json arbitrarily (depending on which Task was executed last).