#1: You need to configure the ContentDispositionFilter with the file path as exclusion to open the file instead of downloading it. Please add the below config in config.publish run mode so on all publish instance it will be opened while in author it will be downloaded. If you want in author also to open you can add the same configuration to config it self. But it's only required on publish as publish instances is exposed to public.
#2: I see the path is blocked in "publishfarm" farm file. You need to enable access to the file location i.e. /content/mysite/* to allow the file to load. Ideally there will be a redirect set at the dispatcher because the file will be always accessed like www.website.com/robots.txt and it should serve content from the actual location. So you need to apply the below redirect at the dispatcher as well: