orchestrator: ResourceExhausted error messages don't include current vs max counts

Problem

When the orchestrator returns a ResourceExhausted gRPC error, the API logs it as:

WARN  Node exhausted, trying another node  {sandboxId: ..., nodeId: ..., error: ...}

However the error field only carries a generic message with no actionable numbers, making it hard to tell from logs why placement was rejected. There are three ResourceExhausted sites:

1. Max running sandboxes (sandboxes.go)

runningSandboxes := s.sandboxFactory.Sandboxes.Count()
if runningSandboxes >= maxRunningSandboxesPerNode {
    return nil, status.Errorf(codes.ResourceExhausted,
        "max number of running sandboxes on node reached (%d), please retry",
        maxRunningSandboxesPerNode)   // ← max only, current omitted
}

runningSandboxes is right there in scope but isn't included in the error.

2. Too many sandboxes starting (sandboxes.go — TryAcquire branch)

acquired := s.startingSandboxes.TryAcquire(1)
if !acquired {
    return nil, status.Errorf(codes.ResourceExhausted,
        "too many sandboxes starting on this node, please retry")  // ← no counts at all
}

3. Too many sandboxes resuming (utils.go — Acquire / resume branch)

err := s.startingSandboxes.Acquire(ctx, 1)
if err != nil {
    return status.Errorf(codes.ResourceExhausted,
        "too many sandboxes resuming on this node, please retry")  // ← no counts at all
}

Impact

Without current/max values in the error, the "Node exhausted" log line carries no information about how exhausted the node is, making it impossible to distinguish between:

• A node that is truly at capacity (e.g. current=50, max=50)
• A node that hit the semaphore limit for in-flight starts (e.g. current=10, max=10) while many slots remain

This matters during placement: the API retries across nodes when it sees ResourceExhausted, and there is no structured signal to tell whether the whole fleet is saturated or just a subset of nodes.

Fix

• Add Current() int64 and Limit() int64 methods to AdjustableSemaphore in packages/shared/pkg/utils/resizable_semaphore.go
• Update all three ResourceExhausted error messages to include current=N, max=M

Example after fix:

"max number of running sandboxes on node reached: current=50, max=50, please retry"
"too many sandboxes starting on this node: current=10, max=10, please retry"
"too many sandboxes resuming on this node: current=10, max=10, please retry"

Start by reading packages/shared/pkg/utils/resizable_semaphore.go, then inspect the three ResourceExhausted sites in sandboxes.go and utils.go. Add the semaphore count methods and update each error to report current and max values; done means all three messages include those counts.