[Hotfix] Handle missing data (#440)

This PR changes how missing data for reward targets is identified. It
fixes a bug where missing data causes the script to crash.

The issue seems to come from a column without valid values being
misidentified as float column and the missing value is encoded as `NaN`.
This was not identified as missing data since it only checks for `None`.

With this PR, the pandas function `isna` is used to identify missing
data.

A similar approach was used in #435.

Since local tests are still running, I created this as draft PR. I will
remove the draft status once the local run is successful.

---------

Co-authored-by: Haris Angelidakis <[email protected]>

src/fetch/payouts.py

@@ -107,12 +107,12 @@ def from_series(cls, frame: Series) -> RewardAndPenaltyDatum:
         )
         solver = frame["solver"]
         reward_target = frame["reward_target"]
-        if reward_target is None:
+        if pandas.isna(reward_target):
             log.warning(f"Solver {solver} without reward_target. Using solver")
             reward_target = solver
 
         buffer_accounting_target = frame["buffer_accounting_target"]
-        if buffer_accounting_target is None:
+        if pandas.isna(buffer_accounting_target):
             log.warning(
                 f"Solver {solver} without buffer_accounting_target. Using solver"
             )