docs: cross-repo proposal updates + tier1 production-readiness plan

Two planning documents covering the rationale for everything else in
this branch.

cbm-cross-repo-proposal.md updates:
* Tier 1 status now reflects 1c-1f as shipped on this branch (was
  "proposed"); §6 producer-side signal source matrix marks the .NET DI
  and Java/Go factory rows as captured.
* New §5.7 Tier 1g — Contract-package FQN extraction, decomposed into
  1g.1 (producer-side dual emission) + 1g.1b (AST-time pkg extraction)
  + 1g.1c (deterministic collision resolution) + 1g.2 (consumer-side
  FQN derivation via NuGet/Maven cache scan).
* Status block on §5.7 explains why the entire 1g family is deferred
  to a focused follow-on PR rather than landed here: half-shipping
  just 1g.1 produces dormant FQN nodes that don't change matching
  behavior, the snoonu fleet hit zero <service>/<method> collisions
  across 200+ .proto files, and five rounds of adversarial review on
  the 1g.1 prototype kept surfacing edge cases that only matter when
  collisions actually fire.
* §5.8 sub-tier sequencing updated to slot Tier 1g as a single bundled
  follow-on PR.

tier1-extractor-fixes.md (new):
* Documents the four production-readiness gaps that prevented Tier 1c-
  1f from firing on real .NET/Java/Go fleets: parallel-pipeline
  wiring, C# 12 primary-ctor extraction, protobuf rpc → service
  fallback, and graph-UI cross-galaxy passthrough.
* Adversarial-review follow-ups: incremental-pipeline wiring (Gap 5),
  safer project-wide stub-var fallback (Gap 6), and the Gap 7
  cross-package collision warning + service_qn property mitigation.
* Validated end-to-end on the snoonu fleet (11 services): 249
  CROSS_GRPC_CALLS edges total, distributed across gateway:93,
  internal-gateway:129, catalogue:13, booking:8, mainpage:5,
  promocode:1.

Author: sponger94