ci: add test and lint workflows, trigger on main+dev

[yanyihan-xiaomi]

Summary / 概述

• typecheck.yml: expanded branch triggers from dev-only to main + dev
  分支触发从仅 dev 扩展为 main + dev
• test.yml: new workflow — unit tests via bun turbo test (Linux, with turbo cache and concurrency control)
  新增单元测试 workflow
• lint.yml: new workflow — bun lint (oxlint static analysis)
  新增 oxlint 静态检查 workflow
• turbo.json: fix task name opencode#test → @mimo-ai/cli#test (package was renamed but turbo config wasn't updated, causing all 326 opencode tests to be silently skipped)
  修复 turbo 任务名，之前包名改了但 turbo.json 没跟上，导致 326 个测试一直没在跑
• AGENTS.md: fix outdated default branch info
  修正过时的默认分支信息
• Issue templates: rename OpenCode → MiMoCode, remove upstream Discord link
  Issue 模板品牌更名
• src/plugin/mimo.ts: skip openBrowser() in test/CI environments (NODE_ENV=test or CI=true)
  测试/CI 环境跳过打开浏览器，避免阻塞
• test/cli/tui/thread.test.ts: set trust: true to skip interactive workspace trust prompt in tests
  测试中跳过 workspace trust 交互弹窗

Pre-existing test failures (not introduced by this PR) / 已有测试失败（非本 PR 引入）

Fixing the turbo task name exposed 17 pre-existing test failures that were never caught because opencode#test was silently skipped (package renamed to @mimo-ai/cli but turbo.json not updated).

修复 turbo 任务名后暴露了 17 个已有测试失败。这些失败之前从未被发现，因为包名从 opencode 改成了 @mimo-ai/cli 后 turbo.json 没更新，测试一直被静默跳过。

Category / 类别	Tests / 测试	Root cause / 根因
Config change / 配置变更	model limit 200K (F41)	DEFAULT_CONTEXT_WINDOW changed to 1M, test still expects 200K
Config change / 配置变更	MimoFreeAuthPlugin disabled providers	Plugin behavior changed, test not updated
Permission logic / 权限逻辑	general agent denies todo tools	Todo tool now allowed, test expects deny
Permission logic / 权限逻辑	Tool whitelist rejects bash	Whitelist logic changed
Checkpoint refactor / Checkpoint重构	checkpoint started→skipped (×2)	Writer now skips in certain scenarios
Checkpoint refactor / Checkpoint重构	insertRebuildBoundary, renderRebuildContext v3 (×3)	Output format changed (tasks ledger + memory index)
Cache strategy / 缓存策略	llm.stream cache_control placement	Ephemeral marker moved from last tool_result to user text
Retry logic / 重试逻辑	structured-output retry count	retryCount changed from 3 to 4
Timeout/race / 超时竞态	subtask metadata (×2)	5s timeout too tight for CI
Incomplete rename / 重命名未完成	actor terminology	src/tool/actor.ts still contains taskRegistry
Compose skill	dispatch template	Prompt missing subagent_type: "general" reference
Actor spawn	return-format injection (F21)	Message persistence logic changed

These should be addressed in follow-up PRs. / 这些应在后续 PR 中修复。

Test plan / 测试计划

• bun typecheck passes (0 errors)
• bun lint passes (0 errors, warnings only)
• CI typecheck ✅
• CI lint ✅
• CI test: 3235 pass, 17 fail (all pre-existing)