[MRG] remove abundance trimming from the default genome-grist workflow

[ctb]

This PR removes the abundtrim/trim-low-abund steps from the default genome-grist workflow, per #197.

Specifically,

• removes /abundtrim/{sample}.abundtrim.fq.gz from most rule inputs and replaces with /trim/{sample}.trim.fq.gz
• adds new top-level rule abundtrim_reads that does produce /abundtrim/{sample}.abundtrim.fq.gz for all configured samples
• removes khmer from the trim.yml conda environment and puts it under a new conda environment, abundtrim.yml

Benefits -

• the abundtrim step is one of the slowest and most memory-intensive steps when working with large data sets, so removing it makes genome-grist's default workflow much faster.
• it removes a confusing and poorly documented step from genome-grist.
• it takes us one big step closer to removing the khmer package as a required dependency; the only other place khmer is used is in counting k-mers with the unique-kmers script, but we could replace that in the future with some other tool.

Disadvantages -

• this should have no significant disadvantages for genome-grist. The major negative outcome will be that the percentage of the metagenome estimated to be "unknown" will increase by a bit.
• The other disadvantage is that abundance trimming is important for graph-based approaches like spacegraphcats, so people wanting to run sgc downstream of genome-grist will need to run

ref discussion in #197

TODO

• adjust documentation so that it's clear that this trimming step is optional and largely intended for sgc compat
• update docs about putting your own metagenomes in place as /trim/{sample}.trim.fq.gz