[MINOR][BUILD] Decode output of commands during merge script as UTF-8 consistently

### What changes were proposed in this pull request?

In the PR merge script, decode the raw output of subprocess commands like `git` using UTF-8 encoding, consistently.

### Why are the changes needed?

The merge script occasionally fails if run with Python 2 and the output of a command like `git` contains non-ASCII characters. I think this most usually happens when a user name, for example, contains Chinese characters.

This is because the output is decoded according to `sys.getdefaultencoding()`, which is ASCII in Python 2. It's UTF-8 in Python 3, by default.

### Does this PR introduce any user-facing change?

No.

### How was this patch tested?

The change caused a merge that failed before to succeed.

Closes #25991 from srowen/MergePRUTF8.

Authored-by: Sean Owen <[email protected]>
Signed-off-by: HyukjinKwon <[email protected]>

Author: srowen

dev/merge_spark_pr.py

@@ -97,9 +97,9 @@ def fail(msg):
 def run_cmd(cmd):
     print(cmd)
     if isinstance(cmd, list):
-        return subprocess.check_output(cmd).decode(sys.getdefaultencoding())
+        return subprocess.check_output(cmd).decode('utf-8')
     else:
-        return subprocess.check_output(cmd.split(" ")).decode(sys.getdefaultencoding())
+        return subprocess.check_output(cmd.split(" ")).decode('utf-8')
 
 
 def continue_maybe(prompt):