However, to the best of my knowledge, when it comes to mail-server-to-mail-server communications, most emails are still transferred in plain text and not encrypted, making it possible to anybody on the network to read their content.
Correct. SMTP, like HTTP, is plain-text by default.
These days many mail servers support TLS (previously known as SSL) for SMTP. (This includes Gmail.) However, it has the same issues as with HTTP[S]: certificates issued by well-known CAs cost money, and self-signed ones are worthless for protecting against MitM attacks. If your mail server does strict validation of the receiver's certificate (like web browsers do), it might fail to deliver messages to servers that are using self-signed certificates or in-house CAs. If it doesn't, then it cannot be sure that it is talking to the right server and not an impostor.
Also, TLS is a relatively recent addition to SMTP, so even when the recipient's mail server supports TLS, the sender's might not, or it might be disabled (sysadmins are lazy species).
Are there any technologies that give the user some guarantees that his emails are sent securely from end to end ?
Two most common email security standards:
OpenPGP, based on web of trust and free to use. The open-source implementation is GnuPG (for Windows, for Thunderbird), and the original PGP has evolved into the commercial PGP Desktop.
For web-based email clients, FireGPG is a possibility — damn it
S/MIME, based on X.509 infrastructure. Implemented by most desktop clients (including Outlook, Thunderbird, Mail.app). However, relatively unpopular due to the same authority-based structure as TLS/SSL: signed certificates cost money, and self-signed ones cannot be reliably validated.
In both cases, encryption requires that the recipient must be already using the system and have generated/obtained a keypair. (For signing, the sender's keypair is used. The normal usage is to sign and encrypt messages.)
Why not let the user know when encryption is not supported and let him choose if he wants his email to be still delivered ?
Usually submitted messages are queued, and neither the user nor the MTA can know whether the next hop supports TLS or not – until the message is being sent, at which point there is no reliable way to ask the user for confirmation. (They may be AFK, offline, asleep, or a script/program. If I sent the message, I want it to be delivered as soon as possible.)
Besides, with SMTP you never know if the next hop is final or if it will just relay the mail somewhere else. It's not uncommon for a backup MX to be on an entirely different network.
Therefore. end-to-end security is only possible when both sides are using OpenPGP or S/MIME.