The performance issue must be caused by too many I/O requests to
`conda.txt` from fgrep. This inline expansion should work to reduce # of
read to the `conda.txt`.
original performance:
```
% git rev-parse HEAD
4f76be6a12
% time bash -c 'pyenv rehash'
bash -c 'pyenv rehash' 0.05s user 0.02s system 76% cpu 0.089 total
```
previous commit: ==> 4x slower than original
```
% git rev-parse HEAD
4469d51ef7
% time bash -c 'pyenv rehash'
bash -c 'pyenv rehash' 0.06s user 0.03s system 25% cpu 0.358 total
```
with this workaround: ==> almost same as original
```
% git rev-parse HEAD
3ffe91bdbc69220eaecf6e2088229cc27366c3f3
% time bash -c 'pyenv rehash'
bash -c 'pyenv rehash' 0.05s user 0.00s system 68% cpu 0.082 total
```
