output/Source: release the mutex while the filter runs

The filter can take some time to finish, even more so on a weak
machine with a saturated CPU.  By not holding the mutex during that
time, we can reduce PlayerThread latency a lot, because that thread
needs to synchronize a lot with all outputs.

src/output/OutputThread.cxx

@@ -264,7 +264,7 @@ AudioOutput::WaitForDelay()
 bool
 AudioOutput::FillSourceOrClose()
 try {
-	return source.Fill();
+	return source.Fill(mutex);
 } catch (const std::runtime_error &e) {
 	FormatError(e, "Failed to filter for output \"%s\" [%s]",
 		    name, plugin.name);

src/output/Source.cxx

@@ -23,6 +23,7 @@
 #include "filter/FilterInternal.hxx"
 #include "filter/plugins/ReplayGainFilterPlugin.hxx"
 #include "pcm/PcmMix.hxx"
+#include "thread/Mutex.hxx"
 #include "util/ConstBuffer.hxx"
 #include "util/RuntimeError.hxx"
 
@@ -187,7 +188,7 @@ AudioOutputSource::FilterChunk(const MusicChunk &chunk)
 }
 
 bool
-AudioOutputSource::Fill()
+AudioOutputSource::Fill(Mutex &mutex)
 {
 	if (current_chunk != nullptr && pending_tag == nullptr &&
 	    pending_data.IsEmpty())
@@ -203,6 +204,10 @@ AudioOutputSource::Fill()
 	pending_tag = current_chunk->tag;
 
 	try {
+		/* release the mutex while the filter runs, because
+		   that may take a while */
+		const ScopeUnlock unlock(mutex);
+
 		pending_data = pending_data.FromVoid(FilterChunk(*current_chunk));
 	} catch (...) {
 		current_chunk = nullptr;

src/output/Source.hxx

@@ -36,6 +36,7 @@
 
 struct MusicChunk;
 struct Tag;
+class Mutex;
 class Filter;
 class PreparedFilter;
 
@@ -147,10 +148,13 @@ public:
 	 *
 	 * Throws std::runtime_error on error
 	 *
+	 * @param mutex the #Mutex which protects the
+	 * #SharedPipeConsumer; it is locked by the caller, and may be
+	 * unlocked temporarily by this method
 	 * @return true if any input is available, false if the source
 	 * has (temporarily?) run empty
 	 */
-	bool Fill();
+	bool Fill(Mutex &mutex);
 
 	/**
 	 * Reads the #Tag to be processed.  Be sure to call Fill()