fix: align pilota parsing with standard grammar

[Joshuahoky]

Motivation

Current parser behaviour deviates from the specs Thrift IDL grammar
in numerous places, causing the parser to accept inputs that should be rejected or reject inputs that should be valid. Some of the deviations include:

#	Example	Original	Expected
1	struct Demo { 40000: required i32 value }	accepted — codegen emits field id -25536 (40000 as i16) in encode/decode	reject — field ids must fit in i16
2	structFoo { i: i32 a}	accepted — parsed as struct Foo { … } (keyword glued to ident), same behaviour for unionFoo, enumFoo...	reject — no identifier boundary
3	const i32 A = --1	accepted — double negation allowed	reject — grammar allows one optional sign
4	struct S { 1: map<i32; string> a }	accepted — ; treated as separator	reject — map key/value separator is , only
5	namespace @@!! foo.bar	accepted — @@!! treated as valid namespace	reject — scope is * or an identifier
6	const i32 A = +1	rejected	accept — IntConstant ::= ('+'|'-')? Digit+
7	const double A = .5	rejected	accept — integer part is optional (Digit*)

Solution

• Add Components::keyword() (a just() followed by a non-consuming word boundary that also succeeds at end of input) and use it for every reserved word
• Rewrite IntConstant as ('+' | '-')? (0x HexDigit+ | Digit+) with the sign kept inside the matched slice, and convert via try_map, so out-of-range values become parse errors rather than panics and i64::MIN round-trips.
• DoubleConstant's integer part becomes Digit* for the .5 form, with the exponent-only form still requiring at least one digit.
• Field ids now go through IntConstant and are range-checked into i16 with a try_map, so negative ids work and oversized ids are a diagnostic instead of a silent truncation.
• Scope becomes '*' | Identifier (dots allowed, since py.twisted is in real-world use), which also makes a missing scope an error.
• The map parser uses just(",") instead of list_separator.

[codecov]

Codecov Report

❌ Patch coverage is 99.37888% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 68.77%. Comparing base (d8c1fef) to head (b0c92bd).

Files with missing lines	Patch %	Lines
pilota-thrift-parser/src/parser/constant.rs	98.74%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #370      +/-   ##
==========================================
+ Coverage   68.46%   68.77%   +0.31%     
==========================================
  Files          89       89              
  Lines       25924    26190     +266     
==========================================
+ Hits        17748    18012     +264     
- Misses       8176     8178       +2     

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.