Video calling induces very high GPU load before locking up the machine

### Steps to reproduce

1. Start an Element Call
2. Optionally, start a screen share

### Outcome

#### What did you expect?
Normal video call behavior

#### What happened instead?
Upon starting an Element call, CPU load increases up to a point where CPU load drops and GPU load jumps to 100%. After just a few seconds, this will make the video streams laggy (both ways) before eventually ending up "stuck". The machine is hardly responsive during this time. When the GPU load reverts to 0 and CPU load increases, the streams will get back to normal. Disabling outgoing video will make this happen less often (but it'll still consistently happen), while screen sharing will increase the frequency of this behavior.


I'd be happy to debug this further with some pointers.

### Operating system

Windows 10

### Application version

Element version 1.11.101

### How did you install the app?

_No response_

### Homeserver

Synapse

### Will you send logs?

Yes

Start by reproducing an Element Call on Windows 10 with outgoing video enabled, then repeat with screen sharing and collect the promised logs. Done means video streams remain responsive and the call no longer drives GPU or CPU load high enough to make the machine unresponsive.

Written by the indexing model from the issue text.