lsremote fails if the output is bigger than DATA_UPLOAD_MAX_MEMORY_SIZE

[stsewd]

When executing commands, we are sanitizing the output

readthedocs.org/readthedocs/doc_builder/environments.py

Line 160 in 464f20e

self.output = self.sanitize_output(cmd_stdout)

readthedocs.org/readthedocs/doc_builder/environments.py

Lines 206 to 210 in 464f20e

	sanitized = (
	'.. (truncated) ...\n'
	f'Output is too big. Truncated at {allowed_length} bytes.\n\n\n'
	f'{truncated_output}'
	)

but lsremote depends on the output as is. This is a problem for repositories with a lot of branches/tags

ref https://sentry.io/organizations/read-the-docs/issues/3193177679/?alert_rule_id=229005&alert_timestamp=1649941826119&alert_type=email&environment=production&project=148442

We should exclude cutting the output from that command or do this "sanitation" just when saving the command.

[agjohnson]

Raised in chat: I had added this separate implementation of API requests for users that were hitting request limits:

https://github.com/readthedocs/readthedocs.org/blob/main/readthedocs/doc_builder/environments.py#L238-L255

It might be good to get that out of feature flag, or potentially feature flag the truncation for now with the same API_LARGE_DATA feature at least.

[humitos]

This seems duplicated of #8990