Use the libspeexdsp resampler

AFAICT it runs a little slower? but it's fixed point, and has much
better understood audio characteristics.

lib/speexdsp/libspeexdsp/Makefile.am

@@ -0,0 +1,46 @@
+# Disable automatic dependency tracking if using other tools than gcc and gmake
+#AUTOMAKE_OPTIONS = no-dependencies
+
+EXTRA_DIST=echo_diagnostic.m
+
+AM_CPPFLAGS = -I$(top_srcdir)/include -I$(top_builddir)/include/speex -I$(top_builddir) @FFT_CFLAGS@
+
+lib_LTLIBRARIES = libspeexdsp.la
+
+# Sources for compilation in the library
+if BUILD_KISS_FFT
+  FFTSRC=kiss_fft.c _kiss_fft_guts.h kiss_fft.h kiss_fftr.c kiss_fftr.h 
+else
+if BUILD_SMALLFT
+  FFTSRC=smallft.c
+else
+  FFTSRC=
+endif
+endif
+
+libspeexdsp_la_SOURCES = preprocess.c jitter.c mdf.c fftwrap.c filterbank.c resample.c buffer.c scal.c $(FFTSRC)
+
+noinst_HEADERS = 	arch.h 	bfin.h \
+		fixed_arm4.h \
+		fixed_arm5e.h 	fixed_bfin.h 	fixed_debug.h 	\
+		math_approx.h 		misc_bfin.h 	\
+		fftwrap.h \
+	filterbank.h fixed_generic.h os_support.h \
+	pseudofloat.h smallft.h vorbis_psy.h resample_sse.h resample_neon.h
+
+libspeexdsp_la_LDFLAGS = -no-undefined -version-info @SPEEXDSP_LT_CURRENT@:@SPEEXDSP_LT_REVISION@:@SPEEXDSP_LT_AGE@
+libspeexdsp_la_LIBADD = $(LIBM)
+
+if BUILD_EXAMPLES
+noinst_PROGRAMS = testdenoise testecho testjitter testresample testresample2
+testdenoise_SOURCES = testdenoise.c
+testdenoise_LDADD = libspeexdsp.la @FFT_LIBS@
+testecho_SOURCES = testecho.c
+testecho_LDADD = libspeexdsp.la @FFT_LIBS@
+testjitter_SOURCES = testjitter.c
+testjitter_LDADD = libspeexdsp.la @FFT_LIBS@
+testresample_SOURCES = testresample.c
+testresample_LDADD = libspeexdsp.la @FFT_LIBS@ @LIBM@
+testresample2_SOURCES = testresample2.c
+testresample2_LDADD = libspeexdsp.la @FFT_LIBS@ @LIBM@
+endif