Limit attachment batches to 50 files - #14289

Related implementation: Make commands own attachment flag policy

Description

The repeatable --attach flag currently has no per-command value limit. A
single command can therefore begin validating and uploading an arbitrarily
large batch of local files.

This caps each invocation at 50 attachment values in the shared flag resolver.
The check runs after command-owned conflicts and before attachment path
validation, uploads, or issue and pull request updates. It counts images and
videos together, including duplicate or invalid paths.

The six supported issue and pull request commands now document the limit in
their detailed help. The installable gh skill contains the same guidance.

How did you test this change?

I ran the local gh issue create implementation with 51 generated --attach
values that named missing files. It stopped before reporting a missing path and
printed:

`--attach` accepts at most 50 values per command

I also rendered gh issue create --help. The detailed help said that a command
can attach up to 50 files, while the flag table retained its concise file-format
description.

Key points

• The unexported limit is enforced once for all six commands.
• The count combines images and videos and precedes file validation.
• Existing command-owned conflict errors keep their precedence.
• The limit applies only to one process invocation and adds no persistent state.
• Detailed command help carries the limit without expanding the shared flag row.

Notes for reviewers

Start with the shared flag resolution and its 50 and 51 boundary cases. The six
command edits only add the same sentence to detailed help.

The placement preserves the command-owned policy boundary established in
PR #14255.

Authorship and follow-up

Who wrote this:

• A human wrote it.
• An agent wrote it under close human direction.
• An agent wrote it independently, and no human has guided the implementation beyond the initial prompt.

Who answers review comments:

• @BagToad will read and reply directly. Name the account.
• An agent will draft replies and @username will read them before they are posted.
• Nobody has explicitly committed to replying.