feat(parser): add BIGQUERY, DATABRICKS and SNOWFLAKE dialect presets

[fudianchn]

AI disclosure: this change was prepared with AI coding agents, reviewed and revised line by line by me.

What

Three warehouse presets, following up on the four warehouses named in #2512:

CCJSqlParserUtil.parse("SELECT `col` FROM t WHERE a = 'x\\'yz' AND b = 42#24",
        p -> p.withDialect(Dialect.BIGQUERY))
// -> SELECT `col` FROM t WHERE a = 'x\'yz' AND b = 42

preset	"	#	\	adjacent literals
BIGQUERY	string literal	line comment	escape sequences	WHITESPACE
DATABRICKS	string literal	the #2507 operator	escape sequences	WHITESPACE
SNOWFLAKE	quoted identifier	the operator	escape sequences	OFF (alias)

Why

Every row is taken from the vendor docs, and Redshift, the fourth warehouse surveyed, matches the defaults on every switch and gains no entry.

• GoogleSQL, lexical structure: "Both string and bytes literals must be quoted, either with single (') or double (\") quotation marks", # single-line comments, "Backslashes (\) introduce escape sequences", literal chunking where "the literal value is the concatenation of all these parts" across any whitespace ("Quoted identifiers don't concatenate"); identifiers are backticked.
• Spark, literals: the syntax [ r ] { 'char [ ... ]' | "char [ ... ]" } [ ... ], the escape table, "Chains of string literals are coalesced into a single string literal". The Databricks KB adds that double quoted identifiers need the non-default doubleQuotedIdentifiers conf, so the default reading is the string one, the same preset decision as MySQL and its ANSI_QUOTES mode in feat(parser): support double quoted string literals behind allowDoubleQuotedStrings #2513. # is not a Spark comment.
• Snowflake, string and binary data types: "A backslash escape sequence is a sequence of characters that begins with a backslash (\)", the double quote "is used (as needed) for delimiting object identifiers"; no adjacent-literal concatenation documented.
• Redshift: delimited identifiers for ", PG-style -- comments, and no AWS docs sentence found for backslash escapes or adjacent literals (community evidence points to escapes, unverified), so every row stays with the default and no empty preset entry is added.

Sources: GoogleSQL lexical structure, Spark literals, Databricks KB, double quoted identifiers, Snowflake string and binary data types, Redshift names.

How

Enum entries only; withDialect is data driven since #2511 and unchanged. Stacked on #2514 for the WHITESPACE mode, the diff shrinks automatically once it merges.

Testing

CCJSqlParserUtilTest: testDialectPresetsWarehouses (BigQuery backticks + escape + 42#24 swallowed as comment, Databricks double quoted string + escape + # stays the operator, Snowflake escape + # stays the operator), testDoubleQuotedStringsPreset extended (string reading for BigQuery and Databricks, identifier guard for Snowflake), testAdjacentStringLiteralsWhitespace extended (the warehouse presets carry the mode, Snowflake stays the alias reading). All verified failing with a wrong row added or a right row removed (Databricks # on, Snowflake " on, BigQuery # off, WHITESPACE turned into NEWLINE); full suite green.

[manticore-projects]

Nice work, please consider my one comment on the Feature and then I would merge prompt. Thank you much for your contribution!

[fudianchn]

@manticore-projects The comment on the Feature has not arrived on my side: the review still shows as pending (unsubmitted). Could you submit it? I will address it right away.

[manticore-projects]

Sorry for asking, are those 2 PRs not additive or redundant?

[fudianchn]

@manticore-projects Additive by design: #2514 carries the feature, this PR only the presets on top (the first four commits here are #2514's), and the description notes the same. Once #2514 merges, this diff shrinks to the presets commit. The boolean overload landed in #2514, with this PR rebased on it.