vmm: migration: Converge based on user-provided downtime and timeout

Wire the new `downtime_ms`, `timeout_s`, and `timeout_strategy` fields
from `VmSendMigrationData` into the precopy loop, replacing the previous
hard-coded 5-iteration cap.

Each iteration now evaluates three convergence criteria in order:
- no dirty pages remain;
- the estimated final-iteration downtime is within the configured budget
- or the overall migration timeout has elapsed.

On timeout, `TimeoutStrategy::Cancel` aborts and keeps the VM live on
the source, while `TimeoutStrategy::Force` proceeds regardless of the
downtime target. The convergence callback is updated to return a Result
to propagate the cancel error cleanly up the call stack.

With the recent changes [0], it is fairly easy to implement the new
checks and operate on actual metrics.

These changes are inspired by [1] but differ significantly in details.

[0] https://github.com/cloud-hypervisor/cloud-hypervisor/pull/7799
[1] https://github.com/cloud-hypervisor/cloud-hypervisor/pull/7033

Signed-off-by: Philipp Schuster <philipp.schuster@cyberus-technology.de>
On-behalf-of: SAP philipp.schuster@sap.com

vm-migration/src/context.rs

@@ -40,9 +40,9 @@ pub struct MemoryMigrationContext {
     ///
     /// Please note that this ignores any additional migration overhead and
     /// only looks at the memory transfer itself.
-    estimated_downtime: Option<Duration>,
+    pub estimated_downtime: Option<Duration>,
     /// Begin of the memory migration.
-    migration_begin: Instant,
+    pub migration_begin: Instant,
     /// Duration of the memory migration.
     ///
     /// This is only `None` until the last iteration is finished.