[CM] virtual edit timing

[Bill Schottstaedt]

I checked some virtual edit accessor timings (1.0=access 1 sec of sound in 1 sec):

 * 1.7 GHz Pentium:
 *                   float               int
 *    straight     38   0.001         31    0.001 ; 28 20 if buffer
 *    ramp         40   0.001         33    0.001
 *    ptree        86   0.002         84    0.002 ; y * 0.5
 *    ramp2        95   0.002         109   0.003
 *    ramp3        102  0.002         114   0.003
 *    ptreec       133  0.003         131   0.003 ; y * vct-ref
 *    ptree2       168  0.004         159   0.004
 *    xramp        376  0.009         371   0.009
 *    cosine       476  0.012         511   0.012 ; cosine-channel-via-ptree
 *    xramp2       682  0.016         672   0.016
 *    xen          3600 0.087         3650  0.088

so even the xen (completely unoptimized, general) case is running 10X real-time!
There are no glitches in playback, and only slight hesitation in some redisplays --
where the peak-env is about to kick in.  I've been chortling about this all
morning.