Support binary inputs to Image, Audio, Video, and Document types

[bschloss]

Overview

In libraries that depend on banks, like LlamaIndex, the formats for image, audio, video and documents that are passed around are urls, paths, and bytes strings. Banks currently does not support byte string input.

Core Changes

• Support byte string inputs to relevant formats
• Update Document formats to be more compatible with existing LLMs

[coveralls]

coverage: 94.742% (+0.4%) from 94.361%
when pulling dca33fb on bschloss:feat-accept-bytes-for-filters
into 1c8f4e8 on masci:main.

[masci]

LGTM, thanks!