Revisit the method load_table_from_json()

[chalmerlowe]

The method load_table_from_json() shows up in a number of Issues and PRs as a source of problems OR ambiguity.

Examine this code and determine whether it (and the similar methods load_table_from_*) should be refactored to remove issues and concerns and/or improve usability.

Some of the issues/PRs where load_table_from_json() or similar methods are mentioned include:

ISSUES:

Schema and Autodetect

• load_table_from_dataframe should honor default project and dataset. #843
• Autodetect for BQ happening automatically even with schema defined.  #847
• Inability to cast int to string when appending data to table using load_table_from_json #906
• load_table_from_json interpolates string as int #1228
• BigQuery seems to automatically convert STRING to BYTES if STRING > 186 bytes #1563
• ALLOW_FIELD_ADDITION not working #1095
• Allow load_table_from_dataframe to Ignore Extra Schema Fields #1812

Retries

• feat: add job_retry argument to load_table_from_uri #969
• Remove num_retries parameter from load_table_from_*() methods #1071

Null Values

• load_table_from_dataframe does not error out when nan in a required column - Million dollar bug #1692
• bigquery load_table_from_dataframe from string type will show null values #1737

Default Value

• Default values not initialized when job loading data with the Python API #1495

Misc

• some emoji will become wrong character stored in bigquery while using load_table_from_json #864
• disambiguate missing policy tags from explicitly unset policy tags #981
• Unable to load table from dataframe with overlapping index/column name #1543
• Provide json.dumps kwargs to load_table_from_json #1564
• Client.load_table_from_dataframe() sometimes chooses invalid column type #1650
• Support pyarrow.large_* as column type in dataframe upload/ download #1706
• load_table_toDataframe breaks with Arrow list fields when the list is backed by a ChunkedArray. #1808

PRs:

#280
#865
#868
#873
#1097
#1526
#1634
#1698