Proto: keep AI launch cycle when stitching from a predecessor with AI

When stitching from a protocol where AI was already active (such as alpha's predecessor), AI launch cycle should remain unchanged.

This MR checks this in a migration test and fixes an issue in alpha introduced in !15211 (aa910b80)

Note: This issue is not present in Q. (More accurately, it can be triggered in Q only if adaptive_issuance_force_activation=true and we migrate from P to Q, but adaptive_issuance_force_activation is false on mainnet and all testnets, so it is not a problem)

The problem in !15211 (merged) was assuming that the desired behavior was always the one when adaptive_issuance_force_activation = true, which is correct when stitching from Genesis (as in most tests), but not when stitching from a protocol which already has AI. This would have been visible on a test network that does the Q -> alpha activation, but there is no such test network yet.

Manually testing the MR

CI

In particular:

dune exec tezt/tests/main.exe -- -f tezt/tests/protocol_migration.ml

All tests in this file except the last one should fail before the last commit, and succeed after it.

Checklist

• Document the interface of any function added or modified (see the coding guidelines)
• [N/A] Document any change to the user interface, including configuration parameters (see node configuration)
• Provide automatic testing (see the testing guide).
• [N/A (the bug has never been released)] For new features and bug fixes, add an item in the appropriate changelog (docs/protocols/alpha.rst for the protocol and the environment, CHANGES.rst at the root of the repository for everything else).