feature request: add in-memory cache for entry list

Problem

When running large clusters, calling ListEntries on every reconcile is not really practical, especially when multiple controller managers are connected to a single SPIRE Server cluster.

A nested SPIRE Server setup could be one possible solution, but adding a cache mechanism to spire-controller-manager seems much simpler.
(#663 related to this maybe?)

for context, our largest Kubernetes cluster has around 30k pods in a single cluster, and roughly 300k–500k entries are registered in SPIRE Server in total.

Proposed Solution

• implement in-memory caching mechanism to avoid listing all entries when reconciling
• these features might work well if it's used together
  • entry prefixing #325
  • hint field based filtering (no issue/PR but spire server support this and I use this in my forked code)

Start by tracing the ListEntries calls made during reconciliation in spire-controller-manager and review the related entry-prefixing discussion in #325. Compare the proposed cache with the entry-listing and filtering needs described here, including the possible hint field. Done means reconciliation no longer lists all entries on every cycle while retaining correct entry results.