Implement Banded LCS calculation with scalar and SIMD support

Introduces the BandedLCS function for efficient LCS scoring, supporting both scalar and SIMD implementations. Adds comprehensive unit tests, benchmarking utilities, and error handling for sequence constraints.

pkg/obialign/fastlcsegf.go

@@ -2,6 +2,7 @@ package obialign
 
 import (
 	"git.metabarcoding.org/obitools/obitools4/obitools4/pkg/obiseq"
+	log "github.com/sirupsen/logrus"
 )
 
 // FastLCSEGFScoreByte calculates the score of the Longest Common Subsequence (LCS) between two byte slices.
@@ -355,27 +356,35 @@ func FastLCSEGFScore(seqA, seqB *obiseq.BioSequence, maxError int, buffer *[]uin
 	return FastLCSEGFScoreByte(seqA.Sequence(), seqB.Sequence(), maxError, true, buffer)
 }
 
-// FastLCSScore calculates the score of the longest common subsequence between two bio sequences.
+// FastLCSScore calculates the score of the longest common subsequence between two bio sequences,
+// and the length of the shortest alignment achieving it (the distance is alilen - lcs).
 //
-// if maxError > 0, the maximum allowed error between the sequences is maxError.
-// Otherwise, no error checking is done.
-// If the actual number of errors is larger than maxError, -1 is returned for both values.
+// If maxError >= 0, the result is exact when its distance is at most maxError,
+// and (-1, -1) is returned otherwise. If maxError < 0, no limit is applied.
+// See BandedLCS, which does the work.
 //
 // The score matrix is:
-//   - Matching: 1
+//   - Matching: 1 (bases are compared with the IUPAC rules)
 //   - Mismatch or gap: 0
 //
+// The program stops with a fatal error if a sequence is longer than
+// MaxBandedLCSLength.
+//
 // Parameters:
 // - seqA: The first bio sequence.
 // - seqB: The second bio sequence.
 // - maxError: The maximum allowed error between the sequences. If set to -1, no limit is applied.
-// - buffer: A pointer to a uint64 slice to store intermediate results. If nil, a new slice is created.
+// - buffer: Unused, kept for compatibility; the working memory is pooled.
 //
 // Returns:
 // - The score of the longest common subsequence.
 // - The length of the shortest alignment corresponding to the LCS.
 func FastLCSScore(seqA, seqB *obiseq.BioSequence, maxError int, buffer *[]uint64) (int, int) {
-	score, alilen, _ := FastLCSEGFScoreByte(seqA.Sequence(), seqB.Sequence(), maxError, false, buffer)
+	score, alilen, err := BandedLCS(seqA.Sequence(), seqB.Sequence(), maxError)
+	if err != nil {
+		log.Fatalf("FastLCSScore on sequences %s (%d bp) and %s (%d bp): %v (maximum %d bp)",
+			seqA.Id(), seqA.Len(), seqB.Id(), seqB.Len(), err, MaxBandedLCSLength)
+	}
 
 	return score, alilen
 }