Feat: anti scam / scam detection

[Maximus7474]

This PR uses perceptual hashing and hamming distance to determine if an image is considered as a scam image.

Note

Builds on top of #7 and #8

It includes a command add-scam-image restricted to users with manage message permission to add new images, this does require that the command author have the image and passes it along to the bot.

Detection occurs on any message with more then one attachment, it'll only hash the image type files. If any hashes match or are similar to the ones stored it'll act accordingly:

• Member is bannable: bot will "softban" him and purge his messages
• Member is moderatable (bot can timeout): bot will timeout for 24h and delete the message
• Member is untouchable by the bot: simple message deletion

Depending on the outcome, the bot will log the action differently in the action log.