[Feature]: Auto-compress images on upload in What's On event creation

### Summary

Automatically compress/optimize images inline when a user uploads an event image that exceeds the size limit on the What's On event creation page, eliminating the need to use an external tool.

### Problem / Motivation

Currently, when a creator uploads an event image that is too large, they must complete a cumbersome manual loop:

1. Download the original file
2. Try to upload it on the What's On event creation page
3. Discover the file is too heavy and the upload is rejected
4. Navigate to an external image optimizer tool
5. Upload the file there, compress it, and download the result
6. Return to the What's On page and re-upload the compressed file
7. Submit the event

This friction slows down event creators and is a common pain point. Compressing the image automatically (or offering a one-click in-page compression) would significantly improve the experience.

### Priority

Medium

### Target audience

Creators

### Area

What's On

### Proposed Solution

When an uploaded image exceeds the allowed file size, automatically compress it client-side (e.g. using the Canvas API or a library like `browser-image-compression`) before submission — or display an inline "Optimize image" action that compresses and re-queues the file without leaving the page.

### User Flow

1. Creator selects an image file for their event
2. File exceeds the size limit
3. Page automatically compresses the image in-browser (or prompts with a single "Compress & continue" button)
4. Compressed image is used for the upload — no external tool needed
5. Creator continues filling out and submitting the event form

### Acceptance Criteria

- [ ] Images exceeding the size limit are compressed client-side before upload
- [ ] Compressed image quality is acceptable for event thumbnails
- [ ] No external navigation required — the full flow stays on the event creation page
- [ ] User receives clear feedback if compression still cannot bring the file within limits

### Additional Context

Reported by Kim Currier via Slack.

Requested by Braian Mellor via Slack.