A few Amazon downloader fixes

[carljm]

1. I introduced an off-by-one error on the order groups index in Add order_groups option to Amazon scraper #32. This is the same issue fixed in Fix off by one error. #37, which I just now noticed.
2. If the desired order group is already-selected, the wait_for_page_load on selecting that index fails, because nothing changes on selecting the already-selected option. So only actually select the option if it's not already selected.
3. Digital order invoice downloading was broken because the print_url was wrong for digital invoices. The most robust way to handle this is just to use the link href instead of manually constructing a URL which may get out of date. This requires verifying that we actually have the Invoice link instead of another link that contains orderId in the href (e.g. "Order Details"). Currently I do this by checking the link text; I could instead do it by checking some contents of the href (though I'd have to check for different marker link contents for digital vs physical orders). Both approaches could be broken by website changes, but in this case looking for the text "Invoice" seems more reliable than relying on details of the URL format, which apparently already has changed once for digital invoices.

Verified that these changes work for downloading both physical and digital order invoices.

[carljm]

Hi @jbms, been using this for a while and it definitely fixes some significant problems in the Amazon importer. Any changes you'd like to see before merge?