perf: add benchmark, optimize

Realized about a 10x speedup.

- add benches/million-updates.rs
- optimization: use rand instead of nanorand (because of thread_rng() initialization being muchf aster)
- update README

Signed-off-by: Stephen Waits <steve@waits.net>

proptest-regressions/height_vector.txt

@@ -9,3 +9,4 @@ cc 38ed1e2c313317809f5469d856d2c1c61e4851c64a156cd1bdd0757fa7c77128 # shrinks to
 cc b7d1b5e9fe3e33066b583c4532865373179972e9b5786ecc62438f23bc939e87 # shrinks to x = -0.0, y = 0.0, h = -3.448118e-33, m = -0.0
 cc 3c4fc4f21c82ffe5bcbc740ca9ab598e0894226acb86a3046a3a24cbbcdbf8e6 # shrinks to x = 0.0, y = 0.0, h = -0.0
 cc d1a551d7eb92bf69e3d0cd0b677baac5aaa5a95d6a016fccf67cf57af71c8b0c # shrinks to x = -5.098216e26, y = 0.0, h = 0.0
+cc 7fa6f2f56f2398bae6804c5f0c0f3f47a7304e505c34f3addb676546cefd97f2 # shrinks to x = -0.0, y = 1.500389e38, h = 2.828499e38