stubborn merge segment after a singleton operator

[tconbeer]

Describe the bug
If an operator ends up on a line by itself (after splitting), we don't try hard enough to merge the segment after that operator into the operator's line, which can result in ugly output (for somewhat pathological input)

To Reproduce
From gitlab's project:

{{column|lower}} || 
ENCRYPT_RAW(
    to_binary('{{ get_salt(column|lower) }}', 'utf-8'), 
    to_binary('{{ env_var("SALT_PASSWORD") }}', 'HEX'), 
    to_binary('416C736F4E637265FFFFFFAB', 'HEX')
)

Expected behavior

{{ column | lower }} || encrypt_raw(
    to_binary('{{ get_salt(column|lower) }}', 'utf-8'),
    to_binary('{{ env_var("SALT_PASSWORD") }}', 'HEX'),
    to_binary('416C736F4E637265FFFFFFAB', 'HEX')
)

Actual behavior

{{ column | lower }} ||
encrypt_raw(
    to_binary('{{ get_salt(column|lower) }}', 'utf-8'),
    to_binary('{{ env_var("SALT_PASSWORD") }}', 'HEX'),
    to_binary('416C736F4E637265FFFFFFAB', 'HEX')
)

Additional context
What is the output of sqlfmt --version?
Commit 88617d2 (v0.10)

[tconbeer]

See also:

https://github.com/tconbeer/gitlab-analytics-sqlfmt/blob/7e002680565326278d4ead72cff9c651f27c7169/transform/snowflake-dbt/models/legacy/bamboohr/rpt/bamboohr_rpt_headcount_aggregation.sql#L88-L89

[tconbeer]

Also need to handle open parens that end up on lines by themselves, e.g., over\n(something long)

https://github.com/tconbeer/gitlab-analytics-sqlfmt/blob/7e002680565326278d4ead72cff9c651f27c7169/transform/snowflake-dbt/models/legacy/bamboohr/sensitive/bamboohr_job_role.sql#L42-L44

[tconbeer]

Also:

https://github.com/tconbeer/rittman_ra_data_warehouse/blob/d8dc7bd1c008ca79f9d09c909734e28a66ef6366/models/integration/int_ad_campaign_performance.sql#L4-L5

[tconbeer]

closed by fd66993

[tconbeer]

closed in #210