Workflow: queue/cache memory estimates and high-memory confirmation

[pecomyint]

Stacked on #185 — review/merge that one first; this diff shows only this PR's change.

Summary

• Each server-queue / collector-cache spinbox shows a 1 MP reference estimate (image and image + float64 Q).
• Starting a consumer whose buffers would retain ≥1 GiB asks for confirmation; larger buffers do not raise sustained throughput.
• The profile save review shows a frame-cache estimate when cache sizes change.

Test plan

• pytest tests/unit passes; ruff check src/dashpva/ clean
• New tests in test_workflow_memory_estimates.py

🤖 Generated with Claude Code

[Osayi-ANL]

Please check out the memory-estimate readout update I added in commit c22acee. It makes the estimate compact, keeps it in muted grey italics, and bolds the calculated memory sizes. If you like the change, please re-request my review and I’ll approve.