Switch index behavior when running out of workers

[dominiklohmann]

📔 Description

This fixes a logic error. To reproduce, add the following assertion on current master:

diff --git a/libvast/src/system/index.cpp b/libvast/src/system/index.cpp
index 2cfd19028..6bc869635 100644
--- a/libvast/src/system/index.cpp
+++ b/libvast/src/system/index.cpp
@@ -197,6 +197,7 @@ bool index_state::worker_available() {
 }

 caf::actor index_state::next_worker() {
+  VAST_ASSERT(worker_available());
   auto result = std::move(idle_workers.back());
   idle_workers.pop_back();
   return result;

Compile, then start a node. Ideally, import lots of data—I tested this by ingesting the whole ICTF data set.

Now, start multiple export commands as background processes that don't have much work to do, e.g.,

repeat 100; do vast export -n 100 null '#type ni ""' &; done

The newly added assertion sometimes triggers.

The change made in this PR fixes the behavior switching.

📝 Checklist

• All user-facing changes have changelog entries.
• The changes are reflected on docs.tenzir.com/vast, if necessary.
• The PR description contains instructions for the reviewer, if necessary.

🎯 Review Instructions

Reason about correct behavior. I've requested @lava as our resident index expert for review.

[lava]

Looks good, reviewed this offline in slack call.