[xingzihai]

Overview:
Normalize the key delimiter to lowercase at construction so that nested key lookups work when the delimiter contains uppercase letters. This is a one-line fix in the KeyDelimiter option.

What problem does it solve?:
When a user configures a key delimiter containing uppercase letters (e.g. KeyDelimiter("<SEPARATOR>")), nested key lookups silently fail. This is because all keys are lowercased before being split on the delimiter (strings.ToLower(key) at viper.go:716, then strings.Split(lcaseKey, v.keyDelim) at viper.go:725/1198/1209), but the delimiter itself is stored raw. A lowercased key like severity<separator>foo never contains the uppercase <SEPARATOR>, so strings.Split yields a single flat element and the nested value is lost.

Closes #2069. This is the same fix as the never-merged PR #2084.

Special notes for a reviewer:

• The fix is minimal: one line in the KeyDelimiter option (v.keyDelim = strings.ToLower(d)).
• All 25 use-sites of keyDelim (Get/Set/SetDefault/AllKeys/AllSettings/searchIndexableWithPathPrefixes/etc.) are fixed by normalizing at the single entry point.
• The default delimiter "." is already lowercase and unaffected.
• Sub() copies the already-normalized value, so no change needed there.
• Regression test TestKeyDelimiterCase verifies that a nested YAML value is correctly resolved with an uppercase delimiter (fails pre-fix, passes post-fix).

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}