fix: pass grammatical context through naturaldelta for i18n case-aware translations

[deepakganesh78]

Fixes #263

Problem

naturaltime() composes naturaldelta() output with format strings like %s ago / %s from now. Languages such as German require different grammatical cases depending on the surrounding preposition (e.g. nominative eine Stunde vs dative einer Stunde after vor). The two-step composition made it impossible for translators to provide case-correct forms because naturaldelta had no knowledge of the tense context in which its result would be used.

Root cause

naturaldelta() uses _ngettext() / _() for all its translation strings. naturaltime() then wraps the result with _("%s ago"). There is no mechanism for translators to supply alternate grammatical forms of the time-unit strings depending on context.

Fix

1. humanize.i18n – added _npgettext() (plural + msgctxt gettext, mirrors the existing _pgettext).
2. naturaldelta() – added an optional context parameter. When set, all internal _() / _ngettext() calls are replaced with _pgettext(context, …) / _npgettext(context, …) so translators can provide context-specific forms via msgctxt entries in .po files.
3. naturaltime() – now passes "naturaltime-past" or "naturaltime-future" as context to naturaldelta().

How translators use this

German translators can now add msgctxt entries like:

`po
msgctxt "naturaltime-past"
msgid "a second"
msgstr "einer Sekunde"

msgctxt "naturaltime-future"
msgid "a second"
msgstr "einer Sekunde"
`

Without matching msgctxt translations, output is identical to before (full backward compatibility).

Validation

• Full test suite: 724 passed, 74 skipped
• ruff check clean on all touched files
• Added regression tests verifying:
  • context parameter produces identical English output (no translation active)
  • _npgettext / _pgettext are called with correct context strings
  • naturaltime passes correct tense context to naturaldelta