I think the answer somewhat depends on what kind and intensity of problems we are talking about.
First, however, some thoughts on who else could deal with this:
The editor and the initial quality check (if one exists) have apparently judged the quality of English to be sufficient for you to review the paper. Thus, you have to expect them not to address this issue by themselves.
Copy editors can and hopefully will fix many issues, but as they are not experts of the subject, they can only do so much (also, some publishers have lousy copy editing). For example, I had copy editors miscorrect a triangulation to the triangulation, as they did not know that the triangulation was not unique in the context. In another example, I know cases where non-native speakers coined new technical terms making suboptimal use of English that however becomes only apparent to somebody who understands the concept described by the term.
Also, copy editors only act after the paper is accepted, which does not help if the quality of English is so bad that it’s impeding the review.
Regarding phrasing your advice, I prefer to avoid explicit suggestions what to do, but rather state what what problems you had and what has to change. You do not know the circumstances under which the manuscript was created and what ways to improve the language are available to the authors. For example, it may be that one of the authors has a good command of English but wasn’t strongly involved in writing the manuscript. Make implicit hints strong rather than subtle though, as the authors may fail to notice the latter, given their English skills.
In addition to ff524’s arguments against explicitly recommending native speakers, I think that recommending a native speaker ignores the importance of understanding the subject matter. Somebody with decent English skills¹ who understands the manuscript is often much more valuable than a professional proofreader who is not from the field.
To get some idea how to phrase the advice, I suggest asking yourself the following questions:
Did you fail to understand significant portions of the paper due to bad English? – If yes, you should definitely indicate this, as at least in the next round of review, you need to understand the whole paper. I suggest (and consider it appropriate) to write something along the lines of the following:
Unfortunately, due to shortcomings in the language of the manuscript, I could not fully assess its quality.
This is an honest statement of facts that does not explicitly tell anybody what to do, but strongly implies that something needs to be done. Moreover, you are implicitly saying that this is not about the scientific quality of the manuscript and that you would like to assess its quality if only you could.
Is the quality of English consistently bad? – I often see manuscripts where you can clearly tell that certain passages were written by different authors. If some of the passages are good, this suggests that their author has a sufficient command of English to revise the rest of the manuscript. As an author, this person is likely best suited for the job. However, it could also be that the authors had only some passages proofread for whatever reason.
Thus an explicit suggestion may be inaccurate or confusing and I would suggest pointing out which passages are problematic and praising the others. This should make it sufficiently clear what to do:
While Sections 1 to 3 were well written, I found it difficult to understand the English of Section 4.
Are there some kinds of mistakes that occur particularly often? – If yes, point them out. For example, some authors tend to use multiple compounds wherever possible without properly hyphenating them, or mix up definite and indefinite articles. (In both these cases, correcting them often requires a deep understanding of the subject matter and thus cannot be done by a copy editor.)
Did the review take significantly longer due to bad English? – If yes, and you expect to have another round of review, you can save some time by remarking on the quality of English now, which I consider appropriate as you are volunteering to review after all. For example you could say:
Due to language mistakes, the manuscript was difficult to read.
If not, and if nothing else is wrong with the language that cannot be addressed on a per-sentence basis, a comment to the editor or a negative rating of the language quality in the editorial system (if it asks you for this) may suffice.
¹ preferably with a native language that is different from the authors’ one