Better integration of SPI with DMA and IRQ. Using only 'End' IRQ. Perf could be improved by using 'Started' IRQ to prepare the next buffer while the current one is beeing sent.

src/drivers/BufferProvider.h

@@ -0,0 +1,11 @@
+#pragma once
+#include <cstddef>
+
+namespace Pinetime {
+  namespace Drivers {
+    class BufferProvider {
+      public:
+        virtual bool GetNextBuffer(uint8_t** buffer, size_t& size) = 0;
+    };
+  }
+}