Remove R&D spike docs from houliston/scheduler branch.

Spike write-ups belong in spike worktrees only; keep progress-guarantee content inline in scheduler docs and TaskQueue header.

Co-authored-by: Cursor <cursoragent@cursor.com>

Author: TrentHouliston

docs/explanation/scheduler.md

@@ -189,7 +189,7 @@ The queues are **lock-free** at the algorithm level: no mutexes, and the system
 - Overflow paths use CAS loops on list pointers.
 - Consumers may spin waiting for a producer's `committed` flag.
 
-The hot-path slot claim via `fetch_add` is wait-free within a non-full block. See `docs/spikes/taskqueue-waitfree-assessment.md` for a detailed progress-guarantee analysis.
+The hot-path slot claim via `fetch_add` is wait-free within a non-full block.
 
 ## Group and sync semantics
 

docs/spikes/taskqueue-waitfree-assessment.md

@@ -46,7 +46,7 @@ namespace threading {
              * retired to a graveyard and deleted when the queue is destroyed. Per-producer FIFO is
              * preserved; cross-producer ordering is not guaranteed.
              *
-             * Progress guarantees (see docs/spikes/taskqueue-waitfree-assessment.md):
+             * Progress guarantees:
              * - Wait-free: slot claim via write.fetch_add and enqueue/dequeue on a non-overflow block
              *   once the slot is committed.
              * - Lock-free but not wait-free: block linking (link_next_block), tail/head CAS, graveyard