MapType read back as array<struct<key,value>> instead of map + DecimalType precision lost on read - regression from 0.43.1

[ye11owSub]

Description

After upgrading spark-bigquery-connector from 0.43.1 to a newer version,
reading data back from BigQuery produces two type regressions.

Regression 1: MapType deserialization broken

When writing a Spark DataFrame with MapType(StringType(), StringType()),
the connector correctly converts it to ARRAY<STRUCT<key STRING, value STRING>>
in BigQuery (expected, since BQ has no native MAP type).

However, on read, version 0.43.1 used to reconstruct the Spark MapType
and return Python dict values. The new version returns
array<struct<key:string,value:string>> - a list of Row objects - instead.

Expected (0.43.1):
tags={'a': 'b', 'c': 'd'}

Actual (new version):
tags=[Row(key='a', value='b'), Row(key='c', value='d')]

Read schema shows tags: array<struct<key:string,value:string>> instead of
tags: map<string,string>.

Regression 2: Decimal precision lost on read

Writing DecimalType(12,2) creates a NUMERIC(12, 2) column in BigQuery.
On read, 0.43.1 returned decimal(12,2). The new version returns decimal(38,9),
losing the original precision and scale, and producing values like
Decimal('123.450000000') instead of Decimal('123.45').

Steps to reproduce

1. Write a Spark DataFrame with a MapType(StringType(), StringType()) column
   and a DecimalType(12,2) column to BigQuery via the connector
2. Read back using spark.read.format("bigquery").load(...)
3. Observe schema and values differ from what was written

Impact

Any code that writes MapType columns and reads them back will observe
silent data shape change. Decimal values with explicit precision will
have incorrect scale on read.