Add support for float64 in audresample.remix()

[hagenw]

Closes #15
Alternative implementation to #16

audresample.remix() works totally fine with other data types than float32, so we don't need to restrict it.

The reason why it was retricted to float32 before was the usage of lib.do_mono_mixdown() from the C library for mixdown, but do we really need to use it?

[codecov]

Codecov Report

Merging #17 (4ad8f8d) into master (9e83740) will not change coverage.
The diff coverage is 100.0%.

Impacted Files	Coverage Δ
audresample/core/api.py	100.0% <100.0%> (ø)

[frankenjoe]

The reason why it was retricted to float32 before was the usage of lib.do_mono_mixdown() from the C library for mixdown, but do we really need to use it?

It was introduced to guarantee identical results with devAIce. But I'm fine with replacing it with a numpy function.

Now that we allow any data type in _check_signal(), did you check what consequences this has on the resample function? Does it work with any data type or do we need to add a check there?

[hagenw]

The reason why it was retricted to float32 before was the usage of lib.do_mono_mixdown() from the C library for mixdown, but do we really need to use it?

It was introduced to guarantee identical results with devAIce. But I'm fine with replacing it with a numpy function.

Before we had a test that compared the output of the library with the numpy implementation I'm using now, compare

audresample/tests/test_remix.py

Lines 16 to 17 in 9e83740

	def mixdown(signal):
	return np.atleast_2d(np.mean(signal, axis=0))

I now reverted it and use the C library in the tests. So I would assume we still guarantee identical results with devAIce.

[frankenjoe]

I now reverted it and use the C library in the tests. So I would assume we still guarantee identical results with devAIce.

nice