Use rattler to install packages

[soapy1]

Description

Use rattler to install packages (instead of conda bulit in transactions).

Some benchmarks with command conda create python numpy scipy imagesize

Using this PR, warmed up cache

 $ hyperfine 'python -m conda create -p /tmp/rattlerinstall-wc python numpy scipy imagesize --json --yes' --warmup 5
Benchmark 1: python -m conda create -p /tmp/rattlerinstall-wc python numpy scipy imagesize --json --yes
  Time (mean ± σ):      4.520 s ±  0.128 s    [User: 3.699 s, System: 1.502 s]
  Range (min … max):    4.320 s …  4.758 s    10 runs

Using this PR, cold cache

$ hyperfine 'python -m conda create -p /tmp/rattlerinstall-cc python numpy scipy imagesize --json --yes' -p 'rm -rf ~/.cache/rattler '
Benchmark 1: python -m conda create -p /tmp/rattlerinstall-cc python numpy scipy imagesize --json --yes
  Time (mean ± σ):     74.917 s ±  1.406 s    [User: 9.532 s, System: 4.980 s]
  Range (min … max):   73.475 s … 78.261 s    10 runs

On the main branch, warm cache

$ hyperfine 'python -m conda create -p /tmp/condainstall python numpy scipy imagesize --json --yes'
Benchmark 1: python -m conda create -p /tmp/condainstall python numpy scipy imagesize --json --yes
  Time (mean ± σ):      7.231 s ±  0.093 s    [User: 6.052 s, System: 1.266 s]
  Range (min … max):    7.101 s …  7.428 s    10 runs

On the main branch, cold cache

$ hyperfine 'python -m conda create -p /tmp/condainstall python numpy scipy imagesize --json --yes' -p 'conda clean --all --yes'
Benchmark 1: python -m conda create -p /tmp/condainstall python numpy scipy imagesize --json --yes
  Time (mean ± σ):     66.399 s ±  1.815 s    [User: 16.974 s, System: 3.447 s]
  Range (min … max):   64.342 s … 69.503 s    10 runs

Some benchmarks with command conda create python numpy scipy imagesize matplotlib pandas scikit-learn

Using this PR, warmed up cache

$ hyperfine 'python -m conda create -p /tmp/rattlerinstall python numpy scipy imagesize matplotlib pandas scikit-learn --json --yes' --warmup 1 
Benchmark 1: python -m conda create -p /tmp/condainstall python numpy scipy imagesize matplotlib pandas scikit-learn --json --yes
  Time (mean ± σ):      6.471 s ±  0.480 s    [User: 5.266 s, System: 3.125 s]
  Range (min … max):    5.702 s …  7.453 s    10 runs

On the main branch, warm cache

$ hyperfine 'python -m conda create -p /tmp/condainstall python numpy scipy imagesize matplotlib pandas scikit-learn --json --yes' --warmup 1 
Benchmark 1: python -m conda create -p /tmp/condainstall python numpy scipy imagesize matplotlib pandas scikit-learn --json --yes
  Time (mean ± σ):     16.879 s ±  2.681 s    [User: 14.371 s, System: 2.508 s]
  Range (min … max):   14.827 s … 22.959 s    10 runs