Describe the pull request

WriteTo can emit a value that its own default reader parses back as a different value, so Load(WriteTo(v)) != v.

writeKeyValue wraps a value only when it contains `/\n (→ """…"""), #/; (→ backtick), or has surrounding whitespace (→ "…"). The comment above it says it covers ", but there is no branch for a value that is itself surrounded by matching quotes. The default reader strips surrounding quotes (hasSurroundedQuote in parser.go), so such a value comes back unquoted:

cfg := ini.Empty()
cfg.Section("").Key("k").SetValue(`"quoted"`)
var buf bytes.Buffer
cfg.WriteTo(&buf)            // k = "quoted"
rc, _ := ini.Load(buf.Bytes())
rc.Section("").Key("k").String() // "quoted"  ->  quoted   (corrupted)

This is one root cause — the writer produces a form the default reader re-interprets — with a few faces:

• values surrounded by " or ' ("quoted", '123', "", "a b", …) are unquoted on read;
• a value that wraps an inner quote in whitespace (_"x"_) breaks the "…" branch, because the reader can't strip a "…" pair that contains another ";
• a value opening with """ is read as a multi-line wrapper.

Fix

Route these through the raw-string (backtick) wrapper the reader already inverts verbatim, so a serialized value is read back byte-for-byte. Plain values, and values that merely contain an interior quote (e.g. "one", "two"), are untouched.

I checked this with a small round-trip census (SetValue → WriteTo → Load, default options) over the value domain — plain, empty, numeric, leading/trailing space, interior/surrounding quotes, #/;/=, `/\n, and """. On main 11 of 42 cases were corrupted; after the change 0 of 42 are, and no previously-correct case changed. The regression test covers the corrupt set plus the no-regression boundary (plain values and interior-quote values must stay unwrapped).

Checklist

• I agree to follow the Code of Conduct by submitting this pull request.
• I have read and acknowledge the Contributing guide.
• I have added test cases to cover the new code.