Bound EPUB metadata before parsing

[dlpwaters]

Summary

• cap META-INF/container.xml at 64 KiB and package OPF XML at 2 MiB from ZIP descriptors before decompression
• independently cap actual member reads at limit + 1, reject descriptor/actual-size mismatches, encrypted members, and unsafe member paths
• use the same bounded reader in both normal EPUB metadata scanning and cover discovery
• cap sidecar metadata.opf files at the same 2 MiB limit
• bump Leaf Reader to 1.1.4 and document the metadata limits

Verification

• all 21 Python tests pass, including malicious oversized container/OPF members and a descriptor/actual-byte mismatch
• omarchy plugin validate .
• qmllint -I /usr/share/omarchy/shell BarWidget.qml Panel.qml ReaderApp.qml
• Python, JavaScript, JSON, starter-book checksum, and EPUB archive checks
• git diff --check

Marketplace submission: omacom/omarchy-plugin-marketplace#1393