Filtering by the presence or absence of a key

[seanpianka]

Based on discussion from #16, a new commit has added support for filtering by the absence or presence of keys in a record.

Here's a example filter which builds a filter that should match these semantics:

"Pass message if key_exists(key) && not(key_matches(key, bad_value)) && not(key_matches(key, ""))".

fn filter<D>(drain: D, level: Level) -> KVFilter<D> where D: Drain {
    KVFilter::new(drain, level)
        .only_pass_on_any_key_present(["err".to_string()].iter())
        .always_suppress_any(Some(
            HashMap::from_iter(
                vec![(
                    "err".to_string(),
                    HashSet::from_iter(vec!["None".to_string(), "".to_string()]),
                )]
            )
        ))
}

To verify the behavior, below is a unit test to verify the above semantics of the filter.

#[cfg(test)]
mod tests {
    use slog::{error, debug, info, Logger, Level, Drain};
    use super::filter;

    #[test]
    fn should_not_log_info_messages() {
        let decorator = slog_term::PlainSyncDecorator::new(slog_term::TestStdoutWriter);
        let drain = slog_term::FullFormat::new(decorator).build().fuse();
        let filter = filter(drain, Level::Debug).fuse();
        let logger = Logger::root_typed(filter, o!());

        // should discard
        info!(logger, "NO: test info");
        info!(logger, "NO: test info"; "count" => 10);
        info!(logger, "NO: test error"; "err" => "None");
        info!(logger, "NO: test error"; "err" => "");
        debug!(logger, "NO: test debug");

        // should log to drain
        info!(logger, "YES: test error"; "err" => "Panic!");
        error!(logger, "YES: test error");
    }
}

The output from this test shows several failure cases, i.e. a record that should be filtered was not.

Mar 10 18:58:03.032 INFO NO: test info
Mar 10 18:58:03.033 INFO NO: test info, count: 10
Mar 10 18:58:03.033 INFO NO: test error, err: None
Mar 10 18:58:03.033 INFO NO: test error, err: 
Mar 10 18:58:03.033 INFO YES: test error, err: Panic!
Mar 10 18:58:03.033 ERRO YES: test error

@przygienda I thought it'd be better to create a new issue for this, rather than inside an unrelated PR.