Apscale

- reduced minsize to pool from 5 to 4 in accordance to usearch documentation with regard to "denoising" individual samples. Only sequences with an abundance of at least 3 in one sample of the dataset are used for OTU clustering / denoising.

- bugfix in OTU / ESV table generation. Now is faster and consumes less memory

- bugfix that stopped apscale from writing the OTU table correctly

- minor bugfix on how np.nan's are replaced in the OTU table

- OTU and ESV tables are not written line by line and therefore consume constantly low memory. This is important for large datasets.

- added an option to dereplicate all samples before adding them to the pooled sequences that are used for clustering. Singletons will still be retained and used for mapping. This greatly reduces the number of spurious OTUs as well as reduces RAM needed for global dereplication. This option can still be omitted by setting the "min size to pool" to 1. Default value is 5.