Fix false invalidation when decoding token User Ids

[Chris-Johnston]

This fixes an issue when trying to validate older bot tokens. The first segment is a base64-encoded string that contains a ulong. The format that discord provides will not contain any padding, while Convert.FromBase64String will throw a FormatException when trying to decode from a string that requires it (string length is not a multiple of 4).

This PR:

• Adds a util method for padding base64 strings + tests
• Pads the first segment of the token before trying to decode it
• Adds a test for tokens that require base64 padding

Original conversation can be found here.

[foxbot]

i'm beginning to question whether or not our token invalidation logic is growing too complicated; however, this is probably a reasonable fix to have on the current codebase, so we'll accept now, reconsider later.

[Chris-Johnston]

Given that it's run into a few problems I agree that it might be best to reconsider, but it seems like it's been somewhat effective at helping people identify when they are passing in invalid tokens.