Resolve benchmark sporadic failure

[seabbs]

See conversation in this PR #358

essentially the issue is that main and some target PRs produce output that doesn't have a judge method despite seeming to work as intended. It is unclear what could be driving this.

PRs that failed

• Issue 246: Standarise outputs #358: Modified a large amount of model code

PRs that worked

• Issue 276: Add information to FAQ #382 Only modified documentation in EpiAware
• Issue 366: Add record expected obs and latent models #383 contained new benchmarks and turing models and worked.

[seabbs]

This seems to have resolved itself and have been a problem with the specific PRs we were investigating. If we start to see this again we should reopen.