Bandwidth filtered is the maximum between the bandwidth
measurements and their mean, not the minimum (#28215)

To speed up the scanner.
RTT is not being used at the moment, set the default to 0.

So that flake8 version 3.6.0 does not give error

It was added to know which variables are in Bytes, but it's not
clear and it should be the default unit when it's not specified.

"measured" could be interpreted as if it was only those relays
that were measured, when i means the relays that were measured and
pass certain restrictions.
"eligible" means the measured relays that would be include in
the bandwidth file.

And catch exception.
This way the tests do not need to be modified to include a
consensus path.

This way the method to measure progress calculate only the
statistics and it is easier to test, removing the logic about the
state file, header stats and number of bw lines returned to this
method instead.

It can be the case that after selecting the results to consider
and before scaling, there is not any. In this case the statistics
should also be written in the header.

So that the method to calculate statistics about the number of
relays measured can be tested without having a tor data directory.

To know whether the min percent of measured relays was reached
before or not, and warn when it was.

since the option to remove the bw file link has been removed.

This reverts commit 5b33d3a3.
Since the option to rm the bw file link has been removed from the
cli.

This reverts commit 199c9274.
We decided that this option is not needed when the header include
statistics about the percentage of relays that failed.

Even for a valid domain that usually resolves, sometimes the local DNS
resolver won't be able to resolve it (duh). Catch the exception that gets
thrown in that instance and fail gracefully by returning an empty list.

and rm condition about whether the min percentage of measured
relays was reached. It will be checked when writing the file.
Also pass the arg rm_link to write.

that are not away from some other X secs.

from some other X secs.
In #27338 it was implemented to do not include relays in the
bandwidth files that does not have at least 2 measurements that
are X secs away from each other.
It was implemented removing any results that were not X secs away
from each other, when all the results should be keep when there is
at least one away from the others, as commented in #28061.