feat: Add SQL support for the NORMALIZE string function

[alexander-beedie]

Closes #20692.

Adds SQL support (and docs) for the NORMALIZE¹ string function; takes advantage of the recently-added normalize expression (see: #20483).

Available forms²:

• NFC: Canonical Decomposition, followed by Canonical Composition.
• NFD: Canonical Decomposition.
• NFKC: Compatibility Decomposition, followed by Canonical Composition.
• NFKD: Compatibility Decomposition.

NORMALIZE(strcol, NFC)
NORMALIZE(strcol, NFD)
NORMALIZE(strcol, NFKC)
NORMALIZE(strcol, NFKD)

If the normalization form is not provided, NFC is used by default.

Example

import polars as pl

pl.DataFrame({
    "txt": [
        "Ｔｅｓｔ",
        "Ⓣⓔⓢⓣ",
        "𝕿𝖊𝖘𝖙",
        "𝕋𝕖𝕤𝕥",
        "𝗧𝗲𝘀𝘁",
    ],
}).sql("SELECT NORMALIZE(txt, NFKC) FROM self").to_series()

# shape: (5,)
# Series: 'txt' [str]
# [
#   "Test"
#   "Test"
#   "Test"
#   "Test"
#   "Test"
# ]

Footnotes

1. PostgreSQL string functions ↩

2. Unicode normalization forms ↩

[codecov]

Codecov Report

Attention: Patch coverage is 75.00000% with 4 lines in your changes missing coverage. Please review.

Project coverage is 79.84%. Comparing base (a0d96f2) to head (a8ac92d).
Report is 18 commits behind head on main.

Files with missing lines	Patch %	Lines
crates/polars-sql/src/functions.rs	75.00%	4 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #20705      +/-   ##
==========================================
+ Coverage   79.03%   79.84%   +0.81%     
==========================================
  Files        1559     1560       +1     
  Lines      221238   221433     +195     
  Branches     2529     2530       +1     
==========================================
+ Hits       174851   176813    +1962     
+ Misses      45806    44038    -1768     
- Partials      581      582       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.