performance improvements

Can pick and choose changes from this, but the important one is avoiding creating additional reqwest clients, because this is slow AF. Here's the flamegraph after a few other modifications (mostly in service of further parallelism and also writing to a string buffer and then printing it at the end so as to keep things ordered correctly) but creating clients per check task.  And after moving to only creating a single reqwest Client and passing around an Arc of it.  The latter is almost 10x shorter in terms of samples on the graph, which roughly correlate to execution time, and is noticeably faster to run.