Fail job gracefully when output name exceeds 255 character DB limit

When discovered dataset designations are very long (e.g., taxonomic names),
the job output association name (__new_primary_file_{name}|{designation}__)
can exceed the VARCHAR(255) column limit, causing StringDataRightTruncation
which corrupts the SQLAlchemy session and triggers PendingRollbackError.

Add name length validation in Job.add_output_dataset and related methods
that raises ValueError before the DB INSERT. Catch the error in both the
normal and extended metadata code paths to fail the job with a clear error
message instead of crashing the transaction session.

Fixes https://github.com/galaxyproject/galaxy/issues/22045

test/functional/tools/discover_long_name.xml

@@ -0,0 +1,13 @@
+<tool id="discover_long_name" name="discover_long_name" version="0.1.0" profile="20.09">
+    <command><![CDATA[
+echo "data" > '${output_name}.txt'
+    ]]></command>
+    <inputs>
+        <param name="output_name" type="text" label="Output name" />
+    </inputs>
+    <outputs>
+        <data format="txt" name="output">
+            <discover_datasets pattern="__name_and_ext__" visible="true" />
+        </data>
+    </outputs>
+</tool>

test/functional/tools/sample_tool_conf.xml

@@ -219,6 +219,7 @@
   <tool file="collection_paired_or_unpaired.xml" />
   <tool file="collection_list_paired_or_unpaired.xml" />
   <tool file="discover_sort_by.xml" />
+  <tool file="discover_long_name.xml" />
   <tool file="discover_default_ext.xml" />
   <tool file="expression_forty_two.xml" />
   <tool file="expression_pick_larger_file.xml" />