Use the len64 function to accelerate PreviousSet and PreviousClear

On an Apple M2 our benchmark produces these results:

BenchmarkBitsetOps/NextSet-8       	 2552749	       468.1 ns/op
BenchmarkBitsetOps/NextClear-8     	 2571921	       466.9 ns/op
BenchmarkBitsetOps/PreviousSet-8   	 1760882	       682.4 ns/op
BenchmarkBitsetOps/PreviousClear-8 	 1763605	       680.0 ns/op

Which greatly improves on the prior results of

    BenchmarkBitsetOps/PreviousSet-8           19930             59924 ns/op
    BenchmarkBitsetOps/PreviousClear-8         19912             61002 ns/op

bitset.go

@@ -594,14 +594,23 @@ func (b *BitSet) PreviousSet(i uint) (uint, bool) {
 	if x >= len(b.set) {
 		return 0, false
 	}
-	for {
-		if b.Test(i) {
-			return i, true
-		}
-		if i == 0 {
-			break
+	w := b.set[x]
+	// Clear the bits above the index
+	w = w & ((1 << (wordsIndex(i) + 1)) - 1)
+	if w != 0 {
+		return uint(x<<log2WordSize) + len64(w) - 1, true
+	}
+	x--
+	// bounds check elimination in the loop
+	if x < 0 {
+		return 0, false
+	}
+	for x >= 0 {
+		w = b.set[x]
+		if w != 0 {
+			return uint(x<<log2WordSize) + len64(w) - 1, true
 		}
-		i--
+		x--
 	}
 	return 0, false
 }
@@ -614,14 +623,27 @@ func (b *BitSet) PreviousClear(i uint) (uint, bool) {
 	if x >= len(b.set) {
 		return 0, false
 	}
-	for {
-		if !b.Test(i) {
-			return i, true
-		}
-		if i == 0 {
-			break
+	w := b.set[x]
+	// Flip all bits and find the highest one bit
+	w = ^w
+	// Clear the bits above the index
+	w = w & ((1 << (wordsIndex(i) + 1)) - 1)
+	if w != 0 {
+		return uint(x<<log2WordSize) + len64(w) - 1, true
+	}
+
+	x--
+	// bounds check elimination in the loop
+	if x < 0 {
+		return 0, false
+	}
+	for x >= 0 {
+		w = b.set[x]
+		w = ^w
+		if w != 0 {
+			return uint(x<<log2WordSize) + len64(w) - 1, true
 		}
-		i--
+		x--
 	}
 	return 0, false
 }

bitset_test.go

@@ -2089,10 +2089,11 @@ func TestWord(t *testing.T) {
 }
 
 func TestPreviousSet(t *testing.T) {
-	v := New(10)
+	v := New(128)
 	v.Set(0)
 	v.Set(2)
 	v.Set(4)
+	v.Set(120)
 	for _, tt := range []struct {
 		index     uint
 		want      uint
@@ -2104,6 +2105,9 @@ func TestPreviousSet(t *testing.T) {
 		{3, 2, true},
 		{4, 4, true},
 		{5, 4, true},
+		{100, 4, true},
+		{120, 120, true},
+		{121, 120, true},
 		{1024, 0, false},
 	} {
 		t.Run(fmt.Sprintf("@%d", tt.index), func(t *testing.T) {
@@ -2116,10 +2120,11 @@ func TestPreviousSet(t *testing.T) {
 }
 
 func TestPreviousClear(t *testing.T) {
-	v := New(10)
+	v := New(128)
 	v.Set(0)
 	v.Set(2)
 	v.Set(4)
+	v.Set(120)
 	for _, tt := range []struct {
 		index     uint
 		want      uint
@@ -2131,6 +2136,9 @@ func TestPreviousClear(t *testing.T) {
 		{3, 3, true},
 		{4, 3, true},
 		{5, 5, true},
+		{100, 100, true},
+		{120, 119, true},
+		{121, 121, true},
 		{1024, 0, false},
 	} {
 		t.Run(fmt.Sprintf("@%d", tt.index), func(t *testing.T) {