Fix out-of-bounds access in Levenshtein edit distance

[yinshiyi]

Fix out-of-bounds access in Levenshtein edit distance by using vector C of size m rather than array. (Vectors have certain extra memory overhead and allow for easier resizing or bounds checking).

Prev C[m] array size is m (0...m-1)
in the loop, C[lact] when lact == m or C[i] when lact == m, will trigger out of bound value leads to oscillation between 2 stable outcome due to neg value < threshold

Example debugging trace data before bug fix
Barcode size = 15, pos 812 trigger a false match due to C[m] out of bound (~50% as neg value) < threshold

pos=808, lact=10, m=14, C[lact]=2, threshold=2, C[m]=-9701813
pos=809, lact=11, m=14, C[lact]=2, threshold=2, C[m]=-9701813
pos=810, lact=12, m=14, C[lact]=2, threshold=2, C[m]=-9701813
pos=811, lact=13, m=14, C[lact]=2, threshold=2, C[m]=-9701813
pos=812, lact=14, m=14, C[lact]=-9701812, threshold=2, C[m]=-9701812
pos=813, lact=14, m=14, C[lact]=3, threshold=2, C[m]=3
pos=814, lact=2, m=14, C[lact]=3, threshold=2, C[m]=3
pos=815, lact=2, m=14, C[lact]=2, threshold=2, C[m]=3
pos=816, lact=3, m=14, C[lact]=2, threshold=2, C[m]=3

The other 50%, the out of bound report a large positive value and pos 812 is not matched.

pos=809, lact=11, m=14, C[lact]=2, threshold=2, C[m]=1246254667
pos=810, lact=12, m=14, C[lact]=2, threshold=2, C[m]=1246254667
pos=811, lact=13, m=14, C[lact]=2, threshold=2, C[m]=1246254667
pos=812, lact=14, m=14, C[lact]=3, threshold=2, C[m]=3
pos=813, lact=3, m=14, C[lact]=3, threshold=2, C[m]=3
pos=814, lact=2, m=14, C[lact]=3, threshold=2, C[m]=3
pos=815, lact=2, m=14, C[lact]=2, threshold=2, C[m]=3

After bug fix, C[m] out of bound issue is resolved

pos=809, lact=11, m=14, C[lact]=2, threshold=2, C[m]=3
pos=810, lact=12, m=14, C[lact]=2, threshold=2, C[m]=3
pos=811, lact=13, m=14, C[lact]=2, threshold=2, C[m]=3
pos=812, lact=14, m=14, C[lact]=3, threshold=2, C[m]=3
pos=813, lact=3, m=14, C[lact]=3, threshold=2, C[m]=3
pos=814, lact=2, m=14, C[lact]=3, threshold=2, C[m]=3
pos=815, lact=2, m=14, C[lact]=2, threshold=2, C[m]=3

[yinshiyi]

This bug fix works well for now.
Future work could include further improving lact update loop safely, to avoid all potential out-of-bounds accesses.

restrander/src/ukkonenMatch.cpp

Lines 114 to 123 in 9fe4320

	while (C[lact] > threshold) {
	lact--;
	}
	if (lact == m) {
	if (firstMatchPos == -1 || pos < firstMatchPos) {
	firstMatchPos = pos; // always pick earliest
	}
	} else {
	lact++;
	}

[yinshiyi]

@jakob-schuster , could you please let me know your thoughts when you have time? thanks!