in_s3: enable multiline parsing

[valentinacala]

Is your feature request related to a problem? Please describe.

I noticed me and many users are wondering to be able to recognize logs that are on many lines.

Currently any plugin seemes to satisfy the thing, because:

• the fluentd multiline parsers applys on in_tail input only
• the fluent-plugin-concat plugin works only if the numeber of lines is well defined

Describe the solution you'd like

I propose to embed the multiline parser in the fluent-plugin-s3 as it works for in_tail input

Describe alternatives you've considered

I tried the flluent-plugin-concat plugin, but it ignore the multiline_start_regexp if it is not specified a fixed number of lines

I tried also applying the multiline parser with S3 source, but I got error

Additional context

No response