Add a comment about possible PH parameters for time/space tradeoffs.

Benchmarks of this idea can lead to improvements, by about 10 ns
per match operation, which makes a difference in throughput. But
we keep things simple for now.

src/ph.c

@@ -123,6 +123,16 @@ lg(unsigned n)
 	return (lg);
 }
 
+/*
+ * Note: there's potential here to use parameters so that users can decide
+ * about a time/space tradeoff. It would allow bits to be increased up to
+ * a limit, increasing the size of the hash table and making it sparser
+ * (that's space). Do so until the expected collision rate falls below a
+ * threshold (time), or until an upper bound on the table size is reached.
+ *
+ * An estimate for the expected number of collisions is:
+ * (n * (n - 1)) >> (bits + 1)
+ */
 static uint32_t
 getsz(unsigned n)
 {