Flush proxy-level caching on production deploy

[soxofaan]

When we do a production promotion/deploy, we apparently still serve cached metadata responses from the old instance.

For example, after a production deploy, I got a mix of old and new "capability doc" responses (example looks at backend_version field of capabilities):

$ for i in {0..10}; do curl -s https://openeo.vito.be/openeo/1.0/ | jq '.["backend_version"]'; done
"0.6.4a1.dev20220804+563"
"0.6.4a1.dev20220804+563"
"0.6.4a1.dev20220804+563"
"0.6.4a1.dev20220804+563"
"0.6.4a1.dev20220804+563"
"0.6.4a1.dev20220719+551"
"0.6.4a1.dev20220719+551"
"0.6.4a1.dev20220804+563"
"0.6.4a1.dev20220719+551"
"0.6.4a1.dev20220719+551"
"0.6.4a1.dev20220719+551"

it was only after roughly half an hour that I got consistently the same response

We should include a cache flush in the deploy procedure