Documentation and diagram update for MultimodalQnA phase 3 enhancements

[dmsuehir]

Description

This PR updates:

• The mermaid diagram in the main MultimodalQnA README to more accurately reflect the Microservices being used.
• Adds a blurb about the function of the Microservices that I added to the diagram (whisper and speecht5)
• Based on our blog review meeting, I added a directory structure diagram for docker_compose.
• There were some command that didn't work out of the box. For example, trying to source the set_env.sh file from a different directory gave an error, so changed it to first cd to that directory.
• For the proxy environment vars, the set_env.sh files were doing things like export http_proxy=${your_http_proxy}, however the main README does not say to set your_http_proxy as an env var (it says to set the real var http_proxy, etc). The README files in docker_compose did mention your_no_proxy so I changed those to just be no_proxy. I checked other examples like ChatQnA and their set_env.sh does not try to set proxy envs, so hopefully these are fine to remove.
• Updated the Gaudi driver version mentioned, because the embedding container is built from the v1.18 base (source) and the tgi-gaudi container is version v2.3.1 which supports synapse 1.19 but our machines have v1.18 and it works there as well.

Issues

RFC: https://github.com/opea-project/docs/blob/main/community/rfcs/24-10-02-GenAIExamples-001-Image_and_Audio_Support_in_MultimodalQnA.md
Issue: opea-project#1549

Type of change

List the type of change like below. Please delete options that are not relevant.

• Others (enhancement, documentation, validation, etc.)

Dependencies

N/A

Tests

Manually tested README commands

[okhleif-10]

Looks much better and cleaner this way. Diagram makes sense. LGTM

[HarshaRamayanam]

Thanks. LGTM

[mhbuehler]

NICE!