Investigate ways to avoid performance impacts of backtraces about snafu HOT 2 OPEN

It would be nice to have some quantitative idea of how expensive generating a backtrace is, as a baseline.

from snafu.

This is exactly how failure handles backtraces and is one of the features I'm really missing from it, you could talk to boats about bringing their backtrace logic into snafu.

One approach I was thinking might make sense is to allow users to have the type of their backtrace type in their error be Option<Backtrace> much like how structopt allows you to make a flag optional by wrapping it in an Option, that way it's configurable on a per usage basis whether or not you want to unconditionally collect backtraces or base it on the RUST_BACKTRACE.

from snafu.