fix: 修复 Skill 上传时 data_id 超长和 Jackson 反序列化超限问题

[Rito-w]

📝 Description

修复 Agent Skill 上传/下载流程中的两个 Bug：

• Issue [Bug]: Agent Skill上传时报错：sorg.springframework.dao.DataIntegrityViolationException: PreparedStatementCallback; Data truncation: Data too long for column 'data_id' at row 1 #278：在 SkillServiceImpl.uploadPackage() 中增加前置校验，上传 ZIP 给 Nacos 前先本地解析资源文件路径，用 NacosAiConfigKeyCodec.encodeSegment() 模拟 Nacos 的编码逻辑，检查编码后长度是否超过 config_info.data_id 的 varchar(256) 限制。超过则返回清晰的 INVALID_PARAMETER 错误提示，避免请求打到 Nacos DB 层才报 DataIntegrityViolationException
• Issue [Bug]: himarket Nacos 客户端反序列化异常：Figma JSON 配置文件超限 (StreamReadConstraints) #279：新增 JacksonConfig 全局配置，通过 StreamReadConstraints.overrideDefaultStreamReadConstraints() 将 maxStringLength 从默认 20M 放宽到 50M，使 Nacos SDK 内部的 ObjectMapper 能正常反序列化包含大体积资源（如 Figma JSON）的 Skill 数据

🔗 Related Issues

Fix #278
Fix #279

✅ Type of Change

• Bug fix (non-breaking change which fixes an issue)

🧪 Testing

• Manual testing completed

手动验证步骤：

1. 构造一个资源文件路径超过 126 字符的 Skill ZIP 包上传，确认返回清晰的 INVALID_PARAMETER 错误（而非 INTERNAL_ERROR）
2. 构造一个路径正常的 Skill ZIP 包上传，确认上传成功不受影响

📋 Checklist

• Code has been formatted (mvn spotless:apply for backend)
• Code is self-reviewed
• Comments added for complex code
• Documentation updated (if applicable)
• No breaking changes (or migration guide provided)
• All CI checks pass

📚 Additional Notes

Issue #278 的根因在 Nacos Server 端：NacosConfigAiResourceStorage.save() 对含 / 的资源路径做全量 hex 编码（enc. + 十六进制），导致长度翻倍。当原始路径超过 126 个 ASCII 字符时，编码后超过 varchar(256) 限制。本 PR 在 HiMarket 端做防御性校验，根本修复需要 Nacos 上游调整 DB schema 或编码策略。

[CLAassistant]

All committers have signed the CLA.

[learnerjohn]

Thanks for the contribution. The fix direction looks reasonable for the Skill upload issues around Nacos data_id length and large Skill resource deserialization.

To make this easier to land against the latest main, we will carry the content of this PR into a new maintainer PR and reference this contribution there.

This PR currently also needs sign-off for the DCO check to pass.

[learnerjohn]

Opened #303 to carry this PR content forward against the latest main. We will continue the merge path there and keep this PR referenced for attribution.

Thanks again for the original contribution.