`vizviewer` parses the entire trace file just to read `file_info`

I noticed vizviewer reads the entire JSON file into memory, parses it into JSON, just to read file_info from it, then throws it away:

https://github.com/gaogaotiantian/viztracer/blob/ede823750456c8bbac6935b948e904772e9c1156/src/viztracer/viewer.py#L313-L321

For my current investigation, report.json is 2.7 gigabytes and it takes 13 seconds for orjson to parse it, which seems unnecessary, since my browser will be parsing it again anyway.

If file_info is crucial for things to work, it would probably be neat to save it separately (e.g. as a report.json.file_info sidecar) while the report file is being saved in the first place.

---

EDIT: this may be related to #308, #533.