feat: add Character and Year converters (#1017)

[BigDataDZ]

Purpose of the pull request

Implements two commonly used JDK 8 compatible converters requested by the community task #1017:
java.lang.Character and java.time.Year. Proposal comments:
#1017 (comment) and the amendment below it
(UUID / YearMonth / Instant were proposed by other volunteers in parallel and are left to
them).

Related: #1017

What's changed?

• CharacterStringConverter (org.apache.fesod.sheet.converters.charconverter): bidirectional
  Character <-> STRING. Reading an empty string yields null; strings longer than one character
  throw (wrapped into ExcelDataConvertException by the framework) instead of being silently
  truncated.
• YearStringConverter (org.apache.fesod.sheet.converters.year): bidirectional Year <->
  STRING, default format uuuu, honors @DateTimeFormat custom patterns and
  GlobalConfiguration.locale.

Both follow the existing *StringConverter pattern (e.g. LocalTimeStringConverter) and are
registered in DefaultConverterLoader under putAllConverter, putWriteConverter and
putWriteStringConverter, so they work on read, xlsx write and CSV write alike.

Each converter ships with JUnit 5 unit tests (11 tests total: type keys, round-trips, default and
custom formats, and error cases).

Checklist

• I have read the Contributor Guide.
• I have written the necessary doc or comment.
• I have added the necessary unit tests and all cases have passed.

[BigDataDZ]

Done in 254a11e — renamed to DateUtils.defaultYearFormat, following the sibling fields
defaultDateFormat / defaultLocalDateFormat, and parseYear / format(Year, ...) now fall
back to it when no explicit format is given, so it works as a global default just like the other
two. Tests re-run green (11/11 targeted, 927/927 full suite).

[bengbengbalabalabeng]

It is recommended to add tests for the following scenarios to cover the complete behavior of YearStringConverter / DateUtils:

• DateUtils#parseYear() / format(Year)
• Locale's impact on formatting（format = "G y"，Locale.CHINA -> "公元 2026"，Locale.US -> "AD 2026"）

[BigDataDZ]

Added in faf2207, covering both layers:

1. DateUtilsTest — direct coverage for DateUtils#parseYear / #format(Year, ...):
   default yyyy parse and format (with null and explicit locales), plus a custom pattern
   (uu).
2. Locale coverage — format(Year.of(2026), "G y", Locale.CHINA) → 公元 2026 and
   Locale.US → AD 2026, mirrored through the converter in
   YearStringConverterTest#convertToExcelDataRespectsLocaleForEraPatterns (a
   GlobalConfiguration with Locale.CHINA / Locale.US and a @DateTimeFormat pattern of
   G y).

Also adjusted the existing custom-format test: the previous MM/yyyy pattern is invalid for a
bare Year (no MonthOfYear field — UnsupportedTemporalTypeException), replaced with
'Y'yyyy → Y2026.

Targeted tests 79/79, full fesod-sheet suite 933/933 green.

[bengbengbalabalabeng]

LGTM.

[nkuprins]

Year is registered with putWriteStringConverter but not with putWriteConverter?

[BigDataDZ]

Good catch @nkuprins — that was an oversight, fixed in 7986ca1:
putWriteConverter(new YearStringConverter()) added alongside putWriteStringConverter.

Without it, a Year field failed on the xlsx write path with
ExcelWriteDataConvertException: Can not find 'Converter' support class Year. (CSV write and
reads were unaffected, which is why the converter-level unit tests did not catch it).

Added a full round-trip regression (yearFieldRoundTripsThroughFesodSheet — a Year field
written to xlsx and read back): red without the registration, green with it. Full fesod-sheet
suite green (934/934).

[bengbengbalabalabeng]

Suggest extending AbstractExcelTest and using its createTempFile(...) helper to cover multiple formats (e.g., XLSX, XLS, CSV). And using RoundTripHelper#writeAndRead(...) to simplify the boilerplate read/write code.

Additionally, it would be better to move this test into a dedicated integration test class rather than mixing it with pure unit tests.

[bengbengbalabalabeng]

Same advice: YearStringConverterTest#yearFieldRoundTripsThroughFesodSheet()

[BigDataDZ]

Both suggestions addressed in 9683bf7:

• The field round trips moved out of the unit test classes into a dedicated integration test,
  CharacterAndYearFieldRoundTripTest — extends AbstractExcelTest, parameterized with
  @ExcelFormatSource (XLSX / XLS / CSV) and built on RoundTripHelper#writeAndRead.
• As a bonus, the parameterization now covers XLS and CSV too — legs the previous
  single-format tests did not exercise.

YearStringConverterTest / CharacterStringConverterTest remain as pure unit tests.
Full fesod-sheet suite green.

[bengbengbalabalabeng]

In read scenarios, xlsx/xls also override matching logic of the kind (JavaType, CellDataType), not just (JavaType, null). Therefore, this comment only describes the override situation on the write path and does not cover the behavior of the read path, which can easily lead to misunderstanding.

My personal suggestion is to remove this description entirely (or revise it into a more comprehensive explanation).