fix: parsing JsonOperator

[izveigor]

Follow on to apache/datafusion#6885.
The main problem is the following: input string starting with characters like @, # in many dialects can have both identifiers and operators.
Because of this, there is ambiguity in grammar.

[coveralls]

Pull Request Test Coverage Report for Build 5507745151

• 58 of 70 (82.86%) changed or added relevant lines in 2 files are covered.
• 3 unchanged lines in 1 file lost coverage.
• Overall coverage decreased (-0.03%) to 86.927%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
tests/sqlparser_common.rs	20	21	95.24%
src/tokenizer.rs	38	49	77.55%

Files with Coverage Reduction	New Missed Lines	%
src/tokenizer.rs	3	88.1%

Totals
Change from base Build 5476034295:	-0.03%
Covered Lines:	15194
Relevant Lines:	17479

---

💛 - Coveralls

[izveigor]

cc @alamb

[alamb]

LGTM @izveigor . Thank you very much for the contribution and I apologize for the delay in review

I have some suggested style changes in 38f1864 that I have made in a follow on PR #923

[alamb]

I verified this test covers the changes in this PR by running this test without the code in this PR and it failed (as expected):

---- parse_json_ops_without_colon stdout ----
thread 'parse_json_ops_without_colon' panicked at 'assertion failed: `(left == right)`
  left: `Ok([Query(Query { with: None, body: Select(Select { distinct: None, top: None, projection: [UnnamedExpr(JsonAccess { left: Identifier(Ident { value: "a", quote_style: None }), operator: HashArrow, right: Identifier(Ident { value: "b", quote_style: None }) })], into: None, from: [], lateral_views: [], selection: None, group_by: [], cluster_by: [], distribute_by: [], sort_by: [], having: None, named_window: [], qualify: None }), order_by: [], limit: None, offset: None, fetch: None, locks: [] })])`,
 right: `Err(ParserError("Expected end of statement, found: >"))`: Parse results with PostgreSqlDialect are different from GenericDialect', src/test_utils.rs:60:21
stack backtrace:

[alamb]

This PR introduced a regression - #926

Here is a PR to fix the regression: #927