feat(korean): introduce coarse POS tag mapping for Hangul multicontext tokenizer

[kahye]

Consolidate fine-grained Sejong/Kiwi POS tags (45 tags) into 17 high-level linguistic categories in `HangulFactorizedTokenizer`:
- `NOUN` (NNG, NNP, NNB)
- `PRON` (NP)
- `NUM` (NR, SN)
- `VERB` (VV, VX)
- `ADJ` (VA)
- `COP` (VCP, VCN)
- `ADV` (MAG, MAJ)
- `DET` (MM)
- `INTJ` (IC)
- `JOSA` (JKS, JKC, JKG, JKO, JKB, JKV, JKQ, JX, JC)
- `EOMI_PRE` (EP)
- `EOMI_FINAL` (EF)
- `EOMI_CONN` (EC)
- `EOMI_MOD` (ETM, ETN)
- `AFFIX` (XPN, XSN, XSV, XSA, XR)
- `PUNCT` (SF, SP, SS, SE, SO, SW, W_URL, W_EMAIL, W_HASHTAG, W_MENTION)
- `FOREIGN` (SL, SH)

### Changes
- Added `POS_TAG_MAP` and updated `POS_TAGS` in `data/template/utils/korean/hangul_factorizer.py`.
- Mapped token tags in `_features()`, `encode_char()`, and `encode_text()`.
- Updated test cases in `data/template/utils/korean/test_hangul_factorizer.py`.
- Updated lane 23 POS values in `data/korean_pos_mc/lane_metadata.json`.

### Testing
- `PYTHONPATH=data/template/utils/korean python3 -m unittest test_hangul_factorizer.py` (all tests passed)