perf: faster execution of build_conda_exclusion_list

1) Got rid of useless call to `cat`, much better to simply use sed with
   file as argument.

2) Got rid of `sort -u`. There is no need to sort the list.
   Additionally, the list `pyenv.d/rehash/conda.d/default.list` only has
   unique entries, and even if you have duplicate entries, the function
   will still work. --> No need for sort nor unique.

3) Further improvement is simple, save a cached
cleaned-list-v1.0 in `conda.d` and simple read from that file instead of
doing `sed`, which must be a more expensive operation than simply
reading from file.

pyenv.d/rehash/conda.bash

@@ -18,13 +18,11 @@ if conda_exists; then
   # `conda_shim` to skip creating shims for those binaries.
   build_conda_exclusion_list() {
     shims=()
-    shopt -s nullglob
-    for shim in $(cat "${BASH_SOURCE%/*}/conda.d/"*".list" | sort -u | sed -e 's/#.*$//' | sed -e '/^[[:space:]]*$/d'); do
+    for shim in $(sed 's/#.*$//; /^[[:space:]]*$/d' "${BASH_SOURCE%/*}/conda.d/default.list"); do
       if [ -n "${shim##*/}" ]; then
         shims[${#shims[*]}]="${shim})return 0;;"
       fi
     done
-    shopt -u nullglob
     eval \
 "conda_shim() {
   case \"\${1##*/}\" in