Retry much more persistently

[pdavies]

Hi there, thanks for creating this package, it's great! I can create an issue if needed but this seemed quicker.

I recently tried to onboard to GarminDB. My data stretches back to 2019, and even though garth has some retry behaviour, it wasn't enough. I found that the download of monitoring data would crash out somewhere during a series of ~7k queries with one error or another, at which point all that downloaded data would be lost and I'd need to start again. Eventually it became clear that although the error rate was low - maybe one every few thousand downloads - it was high enough to have a very small chance of actually completing the process.

One attractive way to deal with this would be to retain a partially downloaded state and resume it on the next attempt. But that sounds like effort! Instead this PR just expands greatly upon the default retries. The choice of parameters is fairly arbitrary. You may want to tweak it. But it successfully unblocked me, and I see little downside to a change such as this. It seems worthwhile to retry many times before throwing away thousands of downloads that have taken hours to retrieve!

[tcgoetz]

Since we are no longer using garth, this change no longer applies. Can you look into whether there are comparable changes for GarminConnect?