Make collapse-detection bitexact.

Jean-Marc's original anti-collapse patch used a threshold on the
 content of a decoded band to determine whether or not it should
 be filled with random noise.
Since this is highly sensitive to the accuracy of the
 implementation, it could lead to significant decoder output
 differences even if decoding error up to that point was relatively
 small.

This patch detects collapsed bands from the output of the vector
 quantizer, using exact integer arithmetic.
It makes two simplifying assumptions:
 a) If either input to haar1() is non-zero during TF resolution
     adjustments, then the output will be non-zero.
 b) If the content of a block is non-zero in any of the bands that
     are used for folding, then the folded output will be non-zero.
b) in particular is likely to be false when SPREAD_NONE is used.
It also ignores the case where mid and side are orthogonal in
 stereo_merge, but this is relatively unlikely.
This misses just over 3% of the cases that Jean-Marc's anti-collapse
 detection strategy would catch, but does not mis-classify any (all
 detected collapses are true collapses).

This patch overloads the "fill" parameter to mark which blocks have
 non-zero content for folding.
As a consequence, if a set of blocks on one side of a split has
 collapsed, _no_ folding is done: the result would be zero anyway,
 except for short blocks with SPREAD_AGGRESSIVE that are split down
 to a single block, but a) that means a lot of bits were available
 so a collapse is unlikely and b) anti-collapse can fill the block
 anyway, if it's used.
This also means that if itheta==0 or itheta==16384, we no longer
 fold at all on that side (even with long blocks), since we'd be
 multiplying the result by zero anyway.

libcelt/vq.c

@@ -106,6 +106,8 @@ static void exp_rotation(celt_norm *X, int len, int dir, int stride, int K, int
       while ((stride2*stride2+stride2)*stride + (stride>>2) < len)
          stride2++;
    }
+   /*TODO: We should be passing around log2(B), not B, for both this and for
+      extract_collapse_mask().*/
    len /= stride;
    for (i=0;i<stride;i++)
    {
@@ -153,7 +155,27 @@ static void normalise_residual(int * restrict iy, celt_norm * restrict X,
    while (++i < N);
 }
 
-void alg_quant(celt_norm *X, int N, int K, int spread, int B, celt_norm *lowband,
+static unsigned extract_collapse_mask(int *iy, int N, int B)
+{
+   unsigned collapse_mask;
+   int N0;
+   int i;
+   if (B<=1)
+      return 1;
+   /*TODO: We should be passing around log2(B), not B, for both this and for
+      exp_rotation().*/
+   N0 = N/B;
+   collapse_mask = 0;
+   i=0; do {
+      int j;
+      j=0; do {
+         collapse_mask |= (iy[i*N0+j]!=0)<<i;
+      } while (++j<N0);
+   } while (++i<B);
+   return collapse_mask;
+}
+
+unsigned alg_quant(celt_norm *X, int N, int K, int spread, int B, celt_norm *lowband,
       int resynth, ec_enc *enc, celt_int32 *seed, celt_word16 gain)
 {
    VARDECL(celt_norm, y);
@@ -165,6 +187,7 @@ void alg_quant(celt_norm *X, int N, int K, int spread, int B, celt_norm *lowband
    celt_word32 sum;
    celt_word32 xy;
    celt_word16 yy;
+   unsigned collapse_mask;
    SAVE_STACK;
 
    celt_assert2(K!=0, "alg_quant() needs at least one pulse");
@@ -308,17 +331,20 @@ void alg_quant(celt_norm *X, int N, int K, int spread, int B, celt_norm *lowband
       normalise_residual(iy, X, N, K, yy, gain);
       exp_rotation(X, N, -1, B, K, spread);
    }
+   collapse_mask = extract_collapse_mask(iy, N, B);
    RESTORE_STACK;
+   return collapse_mask;
 }
 
 
 /** Decode pulse vector and combine the result with the pitch vector to produce
     the final normalised signal in the current band. */
-void alg_unquant(celt_norm *X, int N, int K, int spread, int B,
+unsigned alg_unquant(celt_norm *X, int N, int K, int spread, int B,
       celt_norm *lowband, ec_dec *dec, celt_int32 *seed, celt_word16 gain)
 {
    int i;
    celt_word32 Ryy;
+   unsigned collapse_mask;
    VARDECL(int, iy);
    SAVE_STACK;
 
@@ -332,7 +358,9 @@ void alg_unquant(celt_norm *X, int N, int K, int spread, int B,
    } while (++i < N);
    normalise_residual(iy, X, N, K, Ryy, gain);
    exp_rotation(X, N, -1, B, K, spread);
+   collapse_mask = extract_collapse_mask(iy, N, B);
    RESTORE_STACK;
+   return collapse_mask;
 }
 
 void renormalise_vector(celt_norm *X, int N, celt_word16 gain)