A multisignature scheme would be a signature scheme where several parties all sign the same message. So this is not what we are looking for.
What we are looking for is called an aggregate signature scheme. In such a scheme, we have $\ell$ distinct users, each with their own public key $\mathsf{pk}_i$. Each of those users has a distinct message $m_i$ that they would like to sign.
The goal of the aggregate signature scheme is to create a single short signature that convinces any verifier that each user $i$ did indeed sign their respective message $m_i$.
To do so, there in general exist five algorithms:
- $\mathsf{KGen}(1^n)$: outputs a keypair $(\mathsf{sk_i},\mathsf{pk_i})$.
- $\mathsf{Sign}(\mathsf{sk_i},m_i)$: outputs a signature $\sigma_i$.
- $\mathsf{Vfy}(\mathsf{pk_i},m_i,\sigma_i)$: outputs $1$ if the signature is valid and $0$ otherwise.
- $\mathsf{Agg}((\mathsf{pk_1},\dots,\mathsf{pk_\ell}),(m_1,\dots,m_\ell),(\sigma_1,\dots,\sigma_\ell))$: outputs an aggregate signature $\sigma$.
- $\mathsf{AggVfy}((\mathsf{pk_1},\dots,\mathsf{pk_\ell}),(m_1,\dots,m_\ell),\sigma)$: outputs $1$ if the aggregate signature is valid for all messages and $0$ otherwise.
Standard Correctness applies, i.e. for all messages $m_i$ and all key pairs $(\mathsf{sk}_i,\mathsf{pk}_i) \gets \mathsf{KGen}(1^n)$ it holds that
$$\mathsf{Vfy}(\mathsf{pk_i},m_i,\mathsf{Sign}(\mathsf{sk_i},m_i)) = 1$$
and
$$\mathsf{AggVfy}\left((\mathsf{pk_1},\dots,\mathsf{pk_\ell}),(m_1,\dots,m_\ell),\mathsf{Agg}\left(\begin{array}{c}(\mathsf{pk_1},\dots,\mathsf{pk_\ell}),(m_1,\dots,m_\ell),\\(\mathsf{Sign}(\mathsf{sk_1},m_1),\dots,\mathsf{Sign}(\mathsf{sk_\ell},m_\ell)\end{array}\right)\right) = 1$$
In addition to standard unforgeability we also require unforgeability of aggregated signatures. I.e. we require that for all polynomial time adversaries $\mathcal{A}$ there exists a negligible function $\epsilon(n)$ such that
$$\Pr_{(\mathsf{sk}_1,\mathsf{pk}_1) \gets \mathsf{KGen}(1^n)}\left[\mathsf{AggVfy}\left(\begin{array}{c}(\mathsf{pk_1},\dots,\mathsf{pk_\ell}),\\(m_1,\dots,m_\ell),\sigma\end{array}\right)
:\left(\begin{array}{c}\mathsf{pk}_2,\dots,\mathsf{pk}_\ell,\\m_1,\dotsm_\ell,\sigma\end{array}\right)\gets\mathcal{A}^{\mathsf{Sign}(\mathsf{sk}_1,\cdot)}(\mathsf{pk}_1)\right] \leq \epsilon(n)$$
The first instantiation of such a signature scheme is due to Boneh, Gentry, Lynn, and Shacham in their paper Aggregate and Verifiably Encrypted Signatures from Bilinear Maps. (See specifically Section 3.1.) In their scheme an arbitrary number of signatures can be aggregated in an aggregate signature that consists only of a single group element.
There are many other instantiations of this primitive from different assumptions and with special properties such as the ability to do sequential aggregation.
While some of those schemes have probably been implemented at least benchmarking purposes, I'm not aware of any crypto library implementing aggregate signatures.