fix(parse): keep a leading whitespace-only field at the start of a row

[spokodev]

Problem

A whitespace-only field at the start of a row is dropped or truncated, even though the identical field in any other column is preserved (default options, trim: false):

const csv = require('fast-csv');

csv.parseString(' \n',   { headers: false }) // [[]]         expected [[' ']]      — row dropped
csv.parseString('  \n',  { headers: false }) // [[]]         expected [['  ']]
csv.parseString(' ,b\n', { headers: false }) // [['', 'b']]  expected [[' ', 'b']] — first field corrupted

csv.parseString('a,  ,c', { headers: false }) // [['a', '  ', 'c']]  // a whitespace-only non-first field is fine

058B...\t \t \t... (the existing tab-delimiter test) shows whitespace-only fields are expected to be preserved — it just only holds for non-first columns.

RFC 4180 section 2.4: "Spaces are considered part of a field and should not be ignored." With trim defaulting to false, dropping this whitespace is data loss / a wrong row count.

Cause

RowParser.getStartToken locates the row/field start with scanner.nextNonSpaceToken, which skips leading whitespace, so for " \n" it jumps to the row delimiter (empty row) and for " ,b" it jumps to the , (pushing an empty first column). The leading whitespace — the actual unquoted field content — never reaches the column parser.

Fix

Use scanner.nextCharacterToken for the row start, so leading whitespace flows into NonQuotedColumnParser as field content. The two getters differ only when the next character is whitespace, so any non-whitespace row start is byte-identical. Whitespace before a quoted field is unaffected because ColumnParser.parse still uses nextNonSpaceToken for quote detection ( "x",y still parses as ["x", "y"]).

Verification

• Added RowParser.spec.ts cases for a leading whitespace-only field (' ,b\n' → [' ', 'b'], ' \n' → [' ']). They fail on main and pass with the fix.
• Full @fast-csv/parse suite green (374 tests).
• Round-trip fuzz (2,892 force-quoted records with whitespace-heavy fields) format → parse: 0 failures.

Note (out of scope, flagging for awareness): RowParser.isEmptyRow uses replace(/\s+/g, ''), so ignoreEmpty: true would also discard a legitimate whitespace-only row. Left untouched to keep this PR to one fix.