Fix log parsing undefined variable and duplicate sequence id errors

[dagardner-nv]

Description

• Fixes two related bugs which occur when running the log_parsing example with unexpected data.

• Fix error in LogParsingPostProcessingStage::__get_label_dicts where there is a possibility that the new_label and new_confidence variables are undefined.

• Fix error where the tokenize_text_series method can return a sequence_ids list with duplicate ids when a text series is passed in and one of the values has a length longer than seq_len and truncation=False, this prevents a down-stream error in the inference stage

RuntimeError: Inconsistent ID column. Last element in 'seq_ids' tensor, [1], must not extend beyond last message, [0]

Not sure if this is the best place to catch this error.

Closes #1765

By Submitting this PR I confirm:

• I am familiar with the Contributing Guidelines.
• When the PR is ready for review, new or existing tests cover these changes.
• When the PR is ready for review, the documentation is up to date with these changes.

[dagardner-nv]

/merge