Add test flag to override SYNC_TOLERANCE_EPOCHS for range sync testing (#7030)

Related to #6880, an issue that's usually observed on local devnets with small number of nodes.

When testing range sync, I usually shutdown a node for some period of time and restart it again. However, if it's within `SYNC_TOLERANCE_EPOCHS` (8), Lighthouse would consider the node as synced, and if it may attempt to produce a block if requested by a validator - on a local devnet, nodes frequently produce blocks - when this happens, the node ends up producing a block that would revert finality and would get disconnected from peers immediately.

### Usage

Run Lighthouse BN with this flag to override:

```
--sync-tolerance--epoch 0
```

beacon_node/src/cli.rs

@@ -1509,6 +1509,19 @@ pub fn cli_app() -> Command {
                 .help_heading(FLAG_HEADER)
                 .display_order(0)
         )
+        .arg(
+            Arg::new("sync-tolerance-epochs")
+                .long("sync-tolerance-epochs")
+                .help("Overrides the default SYNC_TOLERANCE_EPOCHS. This flag is not intended \
+                    for production and MUST only be used in TESTING only. This is primarily used \
+                    for testing range sync, to prevent the node from producing a block before the \
+                    node is synced with the network which may result in the node getting \
+                    disconnected from peers immediately.")
+                .hide(true)
+                .requires("enable_http")
+                .action(ArgAction::Set)
+                .display_order(0)
+        )
         .arg(
             Arg::new("gui")
                 .long("gui")