For the minimal deployment in that scenario, you'd have minimally 1x remote collector at the remote site and possible 1x remote collector at the local site to offload the vCenter data collection from the analytics cluster. If monitoring load (objects/metrics) is very low, you can consider running those adapter instances on the analytics clusters.
On the active/active configuration, vR Ops doesn't have a solution other than vROps HA, which only protects against a single node failure. Further, you could consider using SRM to protect the vR Ops analytics clusters to recover at the secondary site.
That being the scenario, you consider doing a vR Ops solution A and vR Ops solution B, with essentially the solution mirrored and centralized at opposite sites (basically 2x deployments that are identical other than node placement).
Most customers will find that active/active simply isn't necessary. Many say they want active/active for an organization availability standard, but again this often isn't justified from an opex/capex perspective. Most don't consider it a BCA. If you do consider it a BCA, consider using SRM as per the above comment. If you consider it a BCA, but don't have SRM, consider the A/B scenario of the solutions.
If you find this or any other answer useful please mark the answer as correct or helpful.
