v1.108.146 — token yield + advisory session budgets

Two reporting-only features that speak the FinOps vocabulary natively.

Advisory session token budget. New session_token_budget config sets an advisory ceiling over response tokens served — the context this server injects into the agent. Past 80% of the limit, every response carries _meta.budget = {limit, spent, state} in-band (approaching at >=80%, over at >=100%), and get_session_stats always reports the block. It never blocks, throttles, or truncates: jCodeMunch is the instrument; hard caps belong to your gateway. 0 (default) disables entirely.

Token yield in get_session_stats. Of the context served this session, how much showed downstream follow-through: served search results later fetched via get_symbol_source/get_context_bundle, or whose file was subsequently edited. rate ships with its components (served_results, followed_through) plus repeated_identical_calls per tool — the agent's redundant context spend, distinct from cache hits. Omitted when nothing rankable was served.

Inline compute only — no new background/persistent/network behavior, no INDEX_VERSION bump, no tool-count or schema change. Suite parity for _meta.budget: jdocmunch-mcp v1.104.0 + jdatamunch-mcp v1.21.0.