Better Error Handling when Bicep installation fails

Using Azure/bicep-deploy@v2 (thus v2.2.0) on a public GitHub worker failed and crashed the workflow with the foloowing as the only error output:

```text
Error: request to https://downloads.bicep.azure.com/releases/latest failed, reason:
```

I re-ran all jobs and it worked again.

Requests:

* Make this step more resilient.
* Try multiple times.
* Fall back to Bicep installed on the worker if present.
* Don't download Bicep on each step? If running multiple bicep-deploy steps, like validate and then what-if, the actions seems to download Bicep each time.
* Maybe don't randomly generate a dir each time, like `Installed Bicep version 0.35.1 to /tmp/bicep-6j9Ypj/bicep`.
* Or cache the downloaded file?

Start with the Bicep installation step used by Azure/bicep-deploy@v2 and reproduce the intermittent download failure described in the issue. Investigate how repeated steps currently download Bicep and create temporary directories. Done means failed downloads are retried, an existing worker installation can be used, and repeated deployment steps avoid unnecessary downloads.

Written by the indexing model from the issue text.