This commit reimplements the core of the "single" mode.  Instead of
doing the detection in the playlist code from the outside, it is moved
to the player thread, which gets a new option called "border_pause".
It will now pause playback exactly at the beginning of the new song,
making the feature more reliable.

Now that the player thread knows what will happen, it can suppress
cross-fading.

Fixes mantis tickets 0003055 and 0003166.

Make sure the player "owns" the next_song object, so nobody else can
free it.

Make sure the decoder "owns" the song object, so nobody else can free
it.

Just in case "song" becomes invalid at some point.

Replaces _current_song().

Rewrite of the pc_get_error_message() function, now using a GError
object instead of the complicated "errored_song" attribute.

Previously, the condition "defined(play_audio_format)" was used to see
if an output device has been opened, but if the device had failed on
startup, an assertion failure could occur.  This patch adds a separate
flag.

Common function that manages "player" attributes after
audio_output_all_open() has returned.

Don't read song.start_ms and song.end_ms, let the player thread manage
this logic instead.

When playing a CUE track, the player thread waited for the decoder to
become ready, and then sent a SEEK command to the beginning of the CUE
track.  If that is near the start of the song file, and the track is
short enough, the decoder could have finished decoding already at that
point, and seeking fails.

This commit makes this initial seek more robust: instead of letting
the player thread deal with the difficult timings, let the decoder API
emulate a SEEK command, and return it to the decoder plugin, as soon
as the plugin finishes its initialization.

These fixes were mostly generated with `codespell' [0] and manually
reviewed.

[0] http://git.profusion.mobi/cgit.cgi/lucas/codespell/

Remove the decoder dependency on player_control.  All player_control
was needed for is to signal the player thread, and we can do that with
a simple GCond as well.

dc_new() allocates the object and returns it.  dc_free() frees it
(replaces dc_deinit()).

Allocate a player_control object where needed, and pass it around.
Each "client" object is associated with a "player_control" instance.

This prepares multi-player support.

When a music_chunk to be crossfaded consists only of a tag,
cross-fading is not possible, and led to an assertion failure.  This
patch just discards those, as if cross-fading was not enabled.

Until the decoder plugin has called decoder_initialized(), the player
may not submit seek commands.  This however could occur with a slow
decoder and a CUE file with a virtual song offset.  This patch adds
another check.

When one song is played twice, and the decoder is working on the
second "instance", but the first should be seeked, the check in
player_seek_decoder() may assume that it can reuse the decoder without
exchanging pipes.  The last thing was the mistake: the pipe pointer
was different, which led to an assertion failure.  This patch adds
another check which exchanges the player pipe.

More abstraction for decoder_control.pipe.

Some abstraction for decoder_control.pipe access.

Replaced all occurrences of g_error() with MPD_ERROR() located in a new header
file 'mpd_error.h'. This macro uses g_critical() to print the error message
and then exits gracefully in contrast to g_error() which would internally call
abort() to produce a core dump.

The macro name is distinctive and allows to find all places with dubious error
handling. The long-term goal is to get rid of MPD_ERROR() altogether. To
facilitate the eventual removal of this macro it was added in a new header
file rather than to an existing header file.

This fixes #2995 and #3007.

Remove cross_fade_apply(), and call pcm_mix() in the output thread,
mixing the chunk and chunk->other together.

Adds mixrampdb and mixrampdelay commands.  Reads MIXRAP_START and
MIXRAMP_END tags from FLAC files and overlaps instead of crossfading.

Previously, tags of the new song being cross-faded in were sent
immediately.  That can cause wrong information being displayed,
because the "previous" song might send its tag at the end again,
overriding the "next" song's tag.  This patch saves & merges the tag
of the next song, and sends it when cross-fading is finished, and the
next song really starts.

Did you ever accidently click "stop" while feeding a radio station?
This option sets the output device to "pause" to disable the "close"
method.  It falls back to "pause" then, which is specific to the
plugin.  Some plugins implement it by feeding silence.

Another quirk fixed: after the last chunk of a song has been played,
the "elapsed_time" variable is set to the chunk's time stamp.  When
the client receives the PLAYER idle event and asks MPD for the current
time stamp, MPD will return the last time stamp of the previous song
when it hasn't played the first chunk of the current song yet.