Required according to RFC 3986:

> An implementation should accept uppercase letters as equivalent to
> lowercase in scheme names

Closes #330

This adds a bit of overhead, but also adds flexibility to the API,
because arbitrary triggers may be invoked from that virtual method
implementation, not just Cond::signal().

The motivation for this is to make the handlers more dynamic, for the
upcoming buffering class utilizing ProxyInputStream.

Yet another potential crash bug fix.

Doesn't make a practical difference - but it's more correct this way.

This way, we have four periods instead of the default of two.  With
only two periods, we don't get woken up often enough, and we
frequently encounter buffer overruns.  With four periods, we have more
time to breathe, and the buffer overruns magically disappear.

This setting is mostly useless for capture devices.  There's no point
in configuring it.

This is necessary because we'll never get woken up again by
epoll_wait() after a buffer overrun recovery, unless we start the PCM
explicitly before returning to the I/O loop.

Copy yet more code from the ALSA output plugin.

Copy some good code from the ALSA output plugin.

Fixes assertion failure in implicit destructor.

Use the snd_pcm_t only in the IOThread, and reuse code that is
well-known to work.