Use realistic central assumptions for dice entropy

[maggo83]

Summary

• base the practical analysis on central bias scenarios of 1% for casino-grade precision dice and 5% for well-made consumer dice
• retain the more pessimistic 10% consumer case as a separate stress scenario rather than the default guide assumption
• document that, under the conservative worst-alignment model, 12-word binary retains 118.990 bits and base-4 retains 121.157 bits for the central consumer scenario
• allow 12- or 24-word phrases for both binary and base-4 in the online, offline, and printable guides
• remove the casino/consumer dice selector and the 12-word binary warning/blocker while keeping the well-made-dice reminder permanently visible
• regenerate the English and German printable guides and update policy tests

Rationale

The previous guide behavior was driven by the deliberately pessimistic 10% consumer-dice stress scenario. The central 5% scenario is a more realistic, though still conservative, assumption for well-made consumer dice and keeps both supported 12-word methods above the 112-bit security objective. The documents continue to state that these scenarios are sensitivity assumptions, not product guarantees.

Validation

• python -m unittest discover -s offline-package/tests -p 'test_*.py' (14 tests, including 200,784 EmBit reference comparisons)
• node offline-package/tests/test_core.js
• online and offline production builds