RFC 2231 continuations percent-decode unencoded segments

## Problem

Parameter continuation assembly percent-decodes every segment once any continuation is encoded. RFC 2231 marks encoding per segment, so literal percent sequences in unencoded segments are incorrectly transformed.

Prepared commit `63c621f8a0722c36d151561edfd4502a385f9a56` decodes only segments carrying the encoded marker while retaining ordering, charset and mixed-continuation behavior.

External models cover encoded/unencoded mixtures, literal percent bytes, Unicode and shuffled segments. The unchanged 1,837-example suite passes on both Rubies; package and Action Mailer integration pass.

CONTRIBUTING requires regression specs, so this report does not open a source-only PR or modify tests.

Search the Ruby mail parser for RFC 2231 parameter continuation assembly, then inspect the existing continuation regression specs and run the 1,837-example suite on both supported Rubies. Done means only segments marked as encoded are percent-decoded, while ordering, charset handling, mixed continuations, and integration behavior remain unchanged.

Written by the indexing model from the issue text.