Report an estimate of remaining steps.

[ehuss]

This adds some reporting to let the user know more about what cargo-bisect-rustc is doing, and provides an estimate of how many more steps it is going to do.

This is not perfect by any means, but seems to provide a decent enough approximation. Some limitations:

• It doesn't consider toolchains that fail to install (unknown_ranges).
• It doesn't differentiate between nightlies and commits. So if you start with nightlies it will tell you N steps left, and then when it starts testing commits, the estimate starts over again.
• It doesn't include the initial start/end check in the estimate. It might be nice, but not trivial to handle.
• It doesn't consider the last check in print_results. Ideally that step should be removed (see Nightly can be checked thrice in certain cases #85).

Fixes #143

[oli-obk]

The behaviour looks good to me. The only impl concern I have is that it distracts from the actual logic. Maybe move the two new fields into a struct and add named methods for all the cases. Then also move cache into the struct and predicate to be a method on the struct.

[ehuss]

I'm having a bit of difficulty fully understanding your suggestion. remaining and estimate need to be recomputed during each loop, so I'm not sure if they benefit from being in a struct.

Can you say more about what you are thinking?

I pushed a change to move the computation into the closure so that it does not distract from the central algorithm. I also removed the hack adjusting the estimate during an Unknown scan, since that estimate will jump around or otherwise be wrong since unknown_ranges is not considered.

Otherwise I'm having a hard time seeing a way to simplify it more.

[oli-obk]

Ah 🤦 I misread the logic. The new thing is much easier to grok, so mission accomplished!