To-do file
----------

A `*vscale' option (up with bsize etc.) allowing you to scale all
volumes to a certain percentage of normal, when the track comes out
too loud or too quiet but you don't want to go through fixing all the
volumes by hand.

Two auto-detect-ish options. One to use the optimum kind of stereo
(real if possible, pseudo if not, and mono on non-stereo devices) and
a more general one to guess suitable sample size and rate. A better
idea might be to have an environment string or rc file to supply opts,
but this is a pain.

Let (standard) samples be loaded from tar file *without having the csf
file being part of it*.

Want a mode that allows for more keyboard-like sounds. Specifically if
a note is played with any already playing, then put all others into
decay part of envelope *if they aren't there already* (will probably
need a flag to test this, which may speed things generally anyway -
think about it, the test 'in decay?' need only be done once I expect).

fix 'retriggered note at end of pattern' problem.

Allow tar files on pipes (i.e. unseekable files), so that the files
can be stored compressed to save disk space. (May also want to allow
invocation of gzip with popen() or similar to make playing compressed
tar files easier.)

The ideas Graham had (in README).

Make csf2tar/tunesam better.

make it output half-a-second or so of silence after an exit when it
was in last_line+1 and was playing to /dev/dsp on Linux, to avoid any
click. :-)
