Fix[Options]: correct handling of packed options

[678098]

Problem

When parsing options, the loop decremented the remaining size by header.words() * WORD_SIZE. For a packed SUB_QUEUE_INFOS option, words() is reinterpreted as the RDA counter - not a length — while the option occupies exactly one word (the header) on the wire. If another option followed a packed one, size was over-decremented, the loop exited early, and the trailing option bytes were misread as payload -> CRC/parse failure and a dropped message. Latent in the common non-fanout case where the packed option is the only one.

Fix

Compute the option's actual on-wire size once (totalSize, packed-aware) and use it consistently for the size decrement and all skip paths, so byte accounting can never desync regardless of the packed flag.

Test

Added OptionsTest.testStreamInPackedInfosThenIds: a packed SUB_QUEUE_INFOS option followed by a SUB_QUEUE_IDS_OLD option. Fails on the old code (trailing option dropped, stream left non-empty); passes with the fix.

[678098]

This size decrement was fully wrong for packed options header.

[678098]

numSkip is equivalent to totalSize - OptionHeader.HEADER_SIZE (in this PR).
I wanted to have exactly 1 expression so I refactored it to use totalSize.