LibAudio: Rescale integer samples correctly in FLAC loader

The FLAC samples are signed, so we need to rescale them not by their bit
depth, but by half of the bit depth. For example, a 24-bit sample
extends from -2^23 to 2^23-1, and therefore needs to be rescaled by 2^23
to conform to the [-1, 1] double sample range.

Userland/Libraries/LibAudio/FlacLoader.cpp

@@ -414,8 +414,7 @@ void FlacLoaderPlugin::next_frame()
 
     VERIFY(left.size() == right.size());
 
-    // TODO: find the correct rescale offset
-    double sample_rescale = static_cast<double>(1 << pcm_bits_per_sample(m_current_frame->bit_depth));
+    double sample_rescale = static_cast<double>(1 << (pcm_bits_per_sample(m_current_frame->bit_depth) - 1));
     dbgln_if(AFLACLOADER_DEBUG, "Sample rescaled from {} bits: factor {:.1f}", pcm_bits_per_sample(m_current_frame->bit_depth), sample_rescale);
 
     m_current_frame_data.clear_with_capacity();