fix(arabic): add missing letters, harakat, tatweel and Arabic-Indic digits

[mohaelmrabet]

Problem

The Arabic ruleset is missing nine letters, including the plain alif ا — the most frequent letter in the language — and the ta marbuta ة, the feminine ending found on a large share of Arabic nouns.

A character with no rule is not transliterated. It falls through to the separator regexp and becomes a dash, so words are cut at arbitrary points and letters silently disappear.

The two existing Arabic test cases document the broken output as if it were expected:

Input	Current output	Expected
هذه هي اللغة العربية	hthh-hy-llgh-laarby	hthh-hy-allghah-alaarbyah
مرحبا العالم	mrhb-laa-lm	mrhba-alaaalm

Two more cases produce results that are arguably worse:

Input	Current output	Note
٢٠٥ (Arabic-Indic digits)	(empty slug)	a page numbered in Arabic digits gets no slug at all
مَدْرَسَة (vocalised)	m-d-r-s	harakat become dashes inside a single word

arabic is in the default ruleset, so this affects every user of Slugify out of the box.

Changes

Only Resources/rules/arabic.json (plus the regenerated DefaultRuleProvider.php and tests). No existing rule is modified — every addition covers a character that previously produced no output, so no current slug changes except the ones that were broken.

Added	Value	Rationale
ا إ آ ٱ	a	alif and its variants; plain alif was absent entirely
ة	ah	ta marbuta, word-final feminine ending
ى	a	alef maksura, pronounced as a final alif
ؤ	o	carried hamza follows its carrier, matching the existing و → o
ئ	a	carried hamza, taking the vowel it actually carries
ء	(removed)	standalone hamza carries no sound of its own
ً ٌ ٍ َ ُ ِ ّ ْ ٰ	(removed)	harakat/tashkeel are vowel marks, not letters
ـ	(removed)	tatweel/kashida is decorative elongation, never lexical
٠–٩	0–9	Arabic-Indic digits

Removing the harakat rather than transliterating them is deliberate: it keeps a vocalised word and its bare spelling slugging identically, which matters for URLs.

مَدْرَسَة  ->  mdrsah
مدرسة      ->  mdrsah
مدرســـة  ->  mdrsah

Cross-checked against ICU

Every rule was compared with ICU's own transliterators (Arabic-Latin and Arabic-Latin/BGN, ICU 74.2) used as an independent oracle. The added letters agree with ICU, with two deliberate exceptions noted below.

Added	This PR	ICU Arabic-Latin	ICU /BGN	Note
ا إ آ	a	a	—	agrees
ٱ	a	(unmapped)	'	ICU leaves it as-is; a is the useful value
ة	ah	t	h	BGN agrees on h; ah restores the vowel the stripped fatha carried
ى	a	y	y	ICU follows the glyph, we follow the pronunciation (على = 'ala)
ئ	a	y	(unmapped)	deliberate, see below
ء	(removed)	ʾ	(removed)	BGN agrees
ـ	(removed)	(removed)	—	agrees
٠–٩	0–9	0–9	0–9	agrees

ICU is not usable as the implementation here: it requires ext-intl, it emits non-ASCII output (ʿ, ā, ẗ) that still needs a Latin-ASCII pass, and Arabic-Latin/BGN leaves characters untransliterated on common words — رئيس → rئيs, فاطمة → fاtmh. It is used here only as a reference to check the table against.

The same cross-check flags six pre-existing rules as diverging from ICU. They are left untouched in this PR, since changing them would alter slugs for current users:

Existing rule	Current	ICU
ج	g	j
ق	k	q
ذ	th	dh
ظ	th	z
ع	aa	ʿ
و	o	w

Happy to open a separate PR for those if you consider a breaking change acceptable for a future major.

Notes

• Transliteration choices were reviewed by a native Arabic speaker.
• php bin/generate-default.php was run, per the contributing guide.
• Test cases added to defaultRuleProvider() per the guide, covering the missing letters, harakat, tatweel and digits.
• Full suite green.