feat: raise agent iteration turn budgets (30/20/50 → 100/100/200)

[fxinfo24]

feat: raise agent iteration turn budgets (30/20/50 → 100/100/200)

Why

Split out from PR #903 (#903) at review request:
these constant bumps are independent of the compaction work and are easier
to review / revert on their own.

Real motivation: in practice, long multi-step agentic runs were hitting the
existing turn ceilings and truncating genuine work before it finished:

• SUBAGENT_DEFAULT_MAX_TURNS 30 → 100 — subagents (no explicit limit)
  aborted mid-task on multi-step work.
• DEFAULT_GOAL_MAX_TURNS 20 → 100 — /goal runs exhausted the budget
  before completing multi-step tasks.
• QueryConfig.max_turns (and frozen twin) 50 → 200 — query/agent loops
  capped mid-run on longer sessions.
• DEFAULT_MAX_TURNS 50 → 200 — interactive session ceiling (shared with
  --max-turns CLI flag), capped mid-task on long runs.

These are 2–4x increases to iteration budgets. They are not required
by the cost-aware compaction path (that feature degrades gracefully when
pricing is unavailable and works at any budget); they simply give real
agentic work room to finish.

Changes

Constant	Before → After	File
SUBAGENT_DEFAULT_MAX_TURNS	30 → 100	src/agent/run_agent.py
DEFAULT_GOAL_MAX_TURNS	20 → 100	src/goals/goals.py
QueryConfig.max_turns	50 → 200	src/query/config.py (line 13)
FrozenQueryConfig.max_turns	50 → 200	src/query/config.py (line 44)
DEFAULT_MAX_TURNS	50 → 200	src/server/agent_server.py

Verification

• All five constants are simple integer literals; no call-site logic changed.
• Syntax-validated the four touched modules (ast.parse).

Note for reviewers

These are pure constant bumps with no behavioral logic. If the larger
budgets are undesirable (e.g. cost/runaway-loop concerns), they revert
cleanly as a unit. Reverting this PR does not affect the compaction work
in #903.

[fxinfo24]

Split out from #903 at review request — these five max-turn constant bumps are independent of the compaction work and are easier to review/revert on their own. Rationale and the full before→after table are in the PR description.

[github-actions]

Test Results

     5 files     997 suites   27m 43s ⏱️
15 374 tests 15 349 ✅ 19 💤 6 ❌
31 022 runs  30 956 ✅ 58 💤 8 ❌

For more details on these failures, see this check.

Results for commit 0e2c51d.

[agentforce314]

LGTM. Thanks!