Fix race condition in async_driver

[knoepfel]

It can happen that the function executed by the async_driver throws an exception before the driver updates its state from "off" to "drive". When this happens, the state of the driver goes from "off" to "park" (due to the exception), and then to "drive". This is an error, and it causes a race condition. Instead, the states should go from "off" to "drive" (on the thread that invokes the driver function) to "park" (which can happen either due to normal completion or an exception).

This PR implements this fix.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

@@            Coverage Diff             @@
##             main     #181      +/-   ##
==========================================
- Coverage   80.93%   80.80%   -0.14%     
==========================================
  Files         115      114       -1     
  Lines        2046     2037       -9     
  Branches      330      328       -2     
==========================================
- Hits         1656     1646      -10     
- Misses        256      257       +1     
  Partials      134      134              

Flag	Coverage Δ
unittests	80.80% <100.00%> (-0.14%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
phlex/utilities/async_driver.hpp	100.00% <100.00%> (ø)

... and 1 file with indirect coverage changes

---

Continue to review full report in Codecov by Sentry.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update a56582b...88c3b5a. Read the comment docs.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[greenc-FNAL]

Is async_driver() accessible from more than one thread (i.e. access to gear_ should be guarded)? It might be possible to define the type of the state enum as std::<atomic> rather than going full guard, but I suspect the call to driver() would need to be grouped under the same guard as gear_

[knoepfel]

Is async_driver() accessible from more than one thread (i.e. access to gear_ should be guarded)? It might be possible to define the type of the state enum as std::<atomic> rather than going full guard, but I suspect the call to driver() would need to be grouped under the same guard as gear_

It turns out that the states enumeration cannot be represented as an std::atomic<int> (https://godbolt.org/z/cbzq3xMj7).
However, gear_ is of type std::atomic<states>. Also, there is only one async_driver per framework program, and access to it is serialized.